McGrath, Kvanvig, and DeRose on Pragmatic Dimensions of Knowledge

There’s been an email exchange between the three of us that makes a nice follow-up to my previous post on pragmatic enroachments into the theory of knowledge, so I’ll post it here to see if anyone wants to pursue the discussion further. I’ll put the longer discussion between Matt and me first, and then put in Keith’s response to Matt as well.

McGrath: Jon, in truth, I regard accepting a practical condition on knowledge as rather unfortunate (since I fully appreciate the appeal of the idea that whether you know depends on your position with respect to the truth — your purely epistemic position). I don’t think, though, that adding anything about the possibility or cost of inquiring further will help, though. Suppose the stakes are high and there’s no real question of inquiring further (somebody threatens you, say, if you inquire further, or maybe it’s just clear you won’t be able to acquire more evidence). This can make a difference to knowledge, it seems. Why? Well, there might be a risky option A and a safe option B. Now, suppose that you know that p. And suppose you know that if p, then A alone will have the best results of your available options. So, it seems you’re in a position to know that B will have worse results than A. How could you rationally do B, then? If we are steadfast in saying that B is what you rationally should do, we’re going to have to give up the claim that you know that B is worse than A, aren’t we? The idea that one should do B even though one knows that B will have worse results than A is very hard to accept, it seems. No appeal to inquiring further will help, I think. That said, it’s also hard to give up the idea that epistemic position determines knowledge-status.

Kvanvig: Matt–the case you describe is interesting. here’s my reaction to it. I don’t think that appeals to further inquiry is what needs to be used to save the purely theoretical nature of knowledge. I’m not sure what will work, but the salience of the possibility of error is surely playing a role in the case you describe. of course, the explanation for why it plays a role adverts to practical considerations, but even so, the appeal to salience itself may be all that is directly needed in the theory.

OK, that depends on a couple of things. first, whether salience can do the work described; and second, whether there is a careful account of salience that will avoid some of the difficulties you’ve pointed out with appeals to salience. The former problem is worse, I think, but to show that it can’t work, we’ll need a case where practical considerations have the effect you describe and yet the person in question fails to take the possibility of error to be unusually salient in the circumstances. When I think of cases of this sort, I don’t find myself tempted to say the person lacks knowledge.

McGrath: Jon, to put it in a nutshell, I am unhappy with the following claims:

1. I should do B even though I know that B will have worse consequences than option A.
2. S should do B even though S knows that B will have worse consequences than option A.

And I think it seems very intuitive to accept the generalization:

If somebody knows an option B has worse consequences than an option A, he shouldn’t do B.

If both 1 and 2 are false, we can’t use salience of error in an invariantist semantics to explain their falsity; we’ll need contextualism. But I don’t think that will work either. There are other alternatives to the falsity hypothesis, like pragmatic implication, of course. But…

Hey, maybe I should just attach an in-progress paper Jeremy and I are writing on such matters — don’t feel you need to give me comments (although they would be appreciated). We go into all this stuff in detail. We haven’t really put together a conclusion, but the rest is probably close to what it will be in the final draft.

Kvanvig: Matt–i got 3 sentences into your paper, and realize where we are separated. I think of an appeal to salience as a move that can preserve the purely epistemic character of justification (if the appeal is handled in terms of an internal defeater). But you characterize epistemicism so that this position is not available.

I’ll read on…

McGrath: Jon, we’re not really separated. I think our first footnote addresses this concern, but maybe we should move it into the body of the text. Really, we don’t look at salience much, but rather practical stuff. We should probably say: traditionally epistemologists have not regarded salience of error as evidence against p (or a defeater).

Kvanvig: After seeing the paper, I don’t think you consider the possibility I suggest. That possibility is that salience is doing all the work, and that salience can be incorporated into epistemicism by adopting a subjective theory of internal defeaters (as in my post on Certain Doubts). Do you see what I mean?

McGrath: Jon, we talked about this in a previous version of the paper, when we had a section on the salience condition SC. I don’t see how salience bears on practical environment, though, unless through a contextualist move. The claim “If S knows that B is worse than A, then S shouldn’t do B” is true whether S is thinking about counterpossibilities or not — I would think. Or better: if it is true, its truth has nothing to do with what S is worrying about.

Kvanvig: Yes, the claim is true, but it won’t do any explanatory work if the antecedent goes false whenever a person starts worrying about possibilities of error. Oops, meant to say “if the antecedent goes falso whenever a person starts worrying about possibilities of error, and in the examples you use to motivate it, that happens. (We should talk about this, cuz I’m having the feeling that we are talking past each other?)

McGrath: We should talk. I think the question is what makes it go false, the fact that the speaker worries or the fact that the subject worries? In first person cases, this is a distinction without a difference. But in a third person cases, there is a real difference. If the antecedent goes false because the speaker worries, we’ve got contextualism. The idea would be that the conditions one must meet to satisfying ‘know’ become more stringent. I think you’d reject this idea. But there are limits to the appeal to salience as a subject-factor. Whether the subject worries or not, if the stakes are high enough, it might be that he ought to worry. If he doesn’t, and so possibilities of error aren’t salient, that doesn’t allow him to “luck out” and have knowledge.

By the way, I think the blog is a huge success. I was skeptical at first, but, wow. I don’t see how you have time to respond to everyone. It’s a full time job.

Kvanvig: OK, now I think we’re on the same page again. Can you think of a case where a person ought to worry, but doesn’t, and hence lucks out in having knowledge?

McGrath: Jon, I think the the general principle is compelling on its face: people shouldn’t do what they know will have worse results than other options available to them. Or consider: people shouldn’t do what they know will have worse results than all other options available to them.

Kvanvig: I agree the principle looks good. Maybe some tiny tinkering from non-consequentialist concerns, but those are details only. My worries are not about the truth of the principle, but the explanation of its truth. So, you claim that the principle is clearly incompatible with epistemicism, but that’s what I’m questioning.

McGrath: For an example, take Cohen’s airport case. I take it that the fact that Mary and John ought to go inquire further about whether the plane stops in Chicago has nothing to do with what’s salient to them. Whether they worry or not, they ought to check further. Now, it doesn’t seem that they could also know that not checking is best. As soon as I think about this, I run a little reductio: if they knew that, then they shouldn’t waste time checking. Think of Smith: he’d mildly prefer that there was no stop in Chicago. But he doesn’t check. Why? He knows that checking is more effort than its worth.

Or take any case in which one knows, say, on the basis of something casual testimony, or a casual induction, of the kind that suffices for knowledge when nothing is really at stake. It can happen that, given the stakes, we think the person shouldn’t do the “risky” option, but should opt for the safe one; this makes us think that she must not know that the safe one is best. Here what the subject is worrying about seems irrelevant. (Oops–should have said: Or take any case in which one knows, say, on the basis of something casual testimony, or a casual induction, of the kind that suffices for knowledge when nothing is really at stake. It can happen that, given the stakes, we think the person shouldn’t do the “risky” option, but should opt for the safe one; this makes us think that she must not know that the risky one is best. Here what the subject is worrying about seems irrelevant.

There’s stuff on this in the paper. Consider a rebuke by a mother: “Oh, Johnny, why are you fussing; you knew I would be gone only a little while.” This counts as a rebuke — regardless of salience of error to Johnny (presumably error was salient to Johnny) — because this knowledge is taken to make it irrational to fuss.

By the way, I don’t think people who ought to worry but don’t luck out and have knowledge. But I think the view you’re attracted to does have this consequence. Mary and John in Cohen’s case, who do worry, don’t know, according to you. But had they not worried, they would.

Kvanvig: Here’s a claim I defend about knowledge (in my new paper on saving coherentism from the lottery paradox): when you know, you have a justification which legitimates the experience of closure with respect to further inquiry. So, if pragmatic factors undermine the legitimacy of closing off further inquiry, they do so by preventing the justification in question from constituting epistemic justification. But since epistemic justification is understood solely in terms of getting to the truth and avoiding error (and not in terms of practical benefits), the role that pragmatic factors play here could be at most indirect.

Actually, I don’t think you disagree with this in the end (maybe I’m wrong, though…) If that’s right, then the only disagreement is over what it takes to have a purely epistemic theory of knowledge. Do you think this take on our discussion is correct?

McGrath: If you want to put this stuff on the blog — together with DeRose’s remarks on “epistemicism” — please feel free. I would be interested in what others had to say. Our discussion has morphed into an interesting back-and-forth. Same for mine with Keith.

DeRose: Hi, Jon. I’ve been in recent e-mail discussions with your colleague, Matt McGrath, and some of what I last wrote to him strikes me as relevant to your post on pragmatic aspects of knowledge. Anyway, while you write of contextualism as being a view that allows for pragmatic encroachment, Matt construes contextualism as claiming the advantage of “epistemicism”:

“epistemicism”: knowledge that p is solely a matter of meeting purely epistemic standards. It’s “the” advantage of contextualism — isn’t it — that it retains commitment to epistemicism.

There’s lots of room for terminololgy to be slipping around among us here, but with that in mind, in case it’s relevant, here’s what I wrote to Matt about contextualism & epistemicism:

Epistemicism: If we have factors somehow divided into “epistemic” and “pragmatic” camps — and I’ll assume we divide those two up about the same — the sense in which contextualism is an epistemicist view is this: Pragmatic factors (of the attributor’s context, which can also, in some contexts, derivatively bring in prag factors of the subject’s context) affect only which standard governs (gives the truth-condition for) a knowledge attribution; but the standards are all such that, given the particular standard selected, whether S satisfies that standard is entirely a matter of epistemic factors. Using somewhat different terms (using “subject factors” for what we’re here calling “epistemic factors”, and using “contextual” and “attributor” factors for what I’m here calling “pragmatic” factors), here’s how I put it in “Contextualism and Knowledge Attributions” (PPR, 1992):

Insofar as we do have *this* belief, that the conditions *for truly saying* that someone
knows do not depend on the sorts of contextual factors we have been discussing, then
contextualism goes against at least one of our beliefs. But it seems that much of the
appeal of this belief derives from the plausibility of the thesis (with which the
contextualist can agree) that whether *we know* something or not does not does not depend
on such factors. The answer to the question, “Does she know?”, in whatever context it is
asked, including a philosophy paper, is determined by facts independent of contextual
factors (or what I have been calling attributor factors). These contextual or attributor
factors affect the content of the question, but once the question is asked with a specific
content, its answer is determined by subject factors, which are precisely the kinds of
factors which *can* very plausibly be thought to affect whether or not the subject knows.


McGrath, Kvanvig, and DeRose on Pragmatic Dimensions of Knowledge — 6 Comments

  1. The issues here involve how to understand epistemicism and the related notion I brought up of an epistemology driven directly by the goal of getting to the truth and avoiding error. Keith thinks of contextualism as committed to epistemicism, as he clarifies it (that knowledge is solely a matter of meeting epistemic standards). Matt clarifies contextualism so that it denies epistemicism, because for Matt, the latter view makes knowledge invariant across subjects who have the same evidence. As such, Matt includes salience about the possibility of error a factor in denying epistemicism.

    The view I’ve been playing with makes salience of the chances of error play a role in the “grounds for doubt” clause of a standard theory of justification, where the justification is epistemic in character. The latter requirement (that the justification be epistemic) implies that the justification has to be strong enough to legitimate the closure experience involved in knowledge, the experience of having concluded that further investigation would undermine present opinion only by revealing misleading information (though I don’t claim that you have to have this experience in order to have knowledge–only that your investigation and/or experience that justifies the belief in question must legitimate this conclusion as well). Both of these features of the view are still love-of-truth-driven, and so do not imply any direct pragmatic element in the theory of knowledge or justification (even though pragmatic issues might explain why the chances of error have become salient in a given case or why the legitimation of the closure conclusion is missing in a given case). In cases where one worries about the chance of error, salience explains the lack of knowledge on this view, and where one does not worry but should worry about the chances of error, one’s justification, I maintain, doesn’t legitimate the closure conclusion. Whether such a view can withstand scrutiny, we’ll see.

    So, we have three different construals of epistemicism at work here. On my construal and Matt’s construal, contextualism is not guaranteed to be a version of epistemicism, while on Keith’s construal, only Matt’s view is a denial of epistemicism (I think).

  2. Epistemicism, as I was thinking of it, is a first order thesis, and so one that doesn’t logically imply either contextualism or invariantism. It states something like: if S and S’ have the same purely epistemic position wrt p, then either both or neither know that p.
    I do think that it’s the main virtue of contextualism over subject sensitive invariantism that it allows one to hold on to epistemicism. Given epistemicism, the contextualist can vindicate our tendency, when we are in high stakes cases or when possibilities of error are salient to us, to make not only first- but also third-person knowledge-denials. (Keith stresses this in recent work). The contextualist can say that people like Mary and John, in Cohen’s airport case, will not only speak the truth in saying “We don’t know,” they will also speak the truth in saying, “And neither does Smith.” The subject sensitive invariantist can’t seem to vindicate these tendencies. She has to say that knowledge that p can vary between subjects who are in the same purely epistemic position wrt p. (The cost for the contextualist, I think, is that she will have to reject general principles like “If someone knows that an option will have the best results, she should do it.”)
    The important question about salience and practical environment, I think, is whether it’s the speaker’s or the subject’s that counts (supposing either counts). If it’s the speaker’s, epistemicism can be preserved by a contextualist maneuver. If it’s the subject’s, I think epistemicism will have to be abandoned.

  3. Matt is right that the view of salience that I adopt does allow a person to know when, had they worried, they wouldn’t know. But that is too direct a consequence of my view for it to move me at all. So let’s look a little deeper to see if there is more here.

    Suppose then with Matt that one doesn’t worry about risks in a given case, but ought to. In the case I’m thinking of, this claim implies the following: suppose a person’s total evidence confirms p, but they ought to have checked further. Matt wants to say that the person doesn’t have knowledge in such a case, and wants to explain the lack of knowledge by appeal to the following principle:

    If somebody knows that not checking has worse consequences than checking, that person should check further.

    Let’s adjust the principle to a decision-theoretic framework amenable to my position, just to see to what extent our differences are about decision theory or about epistemology. Define ‘worse consequences’ to mean: that function on expected utility that appropriately adjusts the expected utility number in conformity with one’s tolerance for risk, where risk-tolerance is understood operationally in terms of where some sort of mental discomfort and stress begins to occur. Call this “risk-adjusted expected utility,” so we’ll say that checking has higher raeu than not checking.

    So suppose you know that checking has higher raeu than not checking, and so ought to check, and hence ought to worry about the chances that you are wrong. But if we are operationalists about risk-tolerance, I don’t see how checking can have higher raeu than not checking, and yet the chances of error not be salient for one. Either the risk tolerance threshold is exceeded, in which case you’ll experience mental discomfort at the risks involved, or it won’t, in which case it is hard to argue that you ought to worry.

    In order to avoid this result, Matt could turn to abnormal psych.; to cases where people defectively fail to worry when a normal person would worry. If that’s what it takes, I think the contextualist is going to win. The case will have to be one where a person is constitutionally cavalier about risks–-rash, as Aristotle would have it. If we say they still should worry, I’m inclined to side with the contextualist-–it is speaker’s context rather than subject’s context that is driving the assessments. That’s because I’m as much a perspectivalist about rational action as I am about justified belief: it is the individual’s own perspective on the world to which ascriptions of rationality or irrationality must answer (though there are many other ways to criticize people who don’t take risks seriously enough–for example, we can criticize them by saying that they are rash, even though rational).

    I think this way of thinking about Matt’s principle will force the disagreements to be disagreements about the decision theory underlying Matt’s principle. Does that generate any progress on the issue?

  4. I think that my intuitions lie with Jon’s here, but I’m not sufficiently clear on things to be sure. So, let me try to summarize, and someone can tell me where I’ve gone wrong. We’ve a question about the semantics of “knows.” In particular, its semantics seem to shift from context to context, one might think (based on recent work of De Rose, Cohen, etc.). What does this mean? We may have a couple pictures. Here is one, or at least somewhere between one and two: I may, when claiming that Jon knows he’s at Mizzou “carry in” a semantics for “knows” by fixing its content in my situation and then ascribing knowledge to him in his situation. When I utter “Jon knows he’s at Mizzou,” the sentence could express a true proposition, and, if Jon is an unusual circumstance, he may utter the sentence “I know I’m at Mizzou” and express a false proposition–he is fixing its content in his situation, not mine.

    So, now the questions are a) what causes such semantic shifts and b) what makes it the case that Jon utters a false proposition? With b), one may say that the truth conditions for *I know I’m am at Mizzou* (when uttered by Jon) are “purely epistemic” (though what this means isn’t totally clear to me, I think I have at least an inchoate grasp). But, pragmatic factors affect which proposition is expressed by “I know I am at Mizzou,” in virtue of affecting the semantics of “knows.” So that’s a). (If you’d like, you can get rid of the first-person indexical and use “Jon” throughout, though I don’t think the indexical here introduces problems.)

    The thesis of a shift in semantics for “knows” is something more interesting than what one deals with in a 101 class, I take it. (So, you can ask how many people know their name, and everyone will say yes; and thirty seconds later you can ask how many people know there is a table at the front of the room and have only half the room say yes. There is a shift in semantics in the following sense: The students are associating a different relation with “knows” in case 1 than they are in case 2. The remedy for this is to try to get everyone on the same page as to roughly which relation “knows” is to be taken to express.)

    With respect to the John and Mary case, whether they ought to worry is totally a function of other justified beliefs they have. (Here I include justified beliefs about classes of beliefs one ought to have, but doesn’t. So mere ignorance doesn’t necessarily get you off the hook here.) I’m not sure if Jon should concede that had they worried, they wouldn’t know. Suppose Mary is going through Xanax withdrawal and is worrying (specifically, relevantly, here), but reasonably believes that this is the cause of her worrying. Is she denied knowledge? I don’t think so; she has a defeater here.

    So, if someone would be so kind:

    a) Can you explain in finitely many sentences where I’ve gone wrong above?
    b) Can I have a Chisholm-style definition of “epistemic” and “pragmatic” as used in these contexts?
    c) Can I have at least a McTaggart-style definition of the above terms, if I can’t have a Chisholm-style one?
    d) Whence the shift in semantics for “knows”?

  5. Matt’s example about Mary and Xanax is very good. That’s right, I shouldn’t say that worrying about the chances of error in itself removes knowledge. Such worrying is a ground for doubt, but it need not be an unrebutted ground for doubt, and in the Mary/Xanax case, her awareness that the worrying is simply a product of withdrawal is a good example of how a defeater can itself be eaten by further information.

    I wish I could give a clear definition of epistemic and pragmatic, but I think Matt’s account of epistemicism is a good starting point: the epistemic is a function of evidence (of course, that won’t work for externalists…!).

    As I see it, there are a couple of issues above. First, is it speaker context of subject context that determines the truth of knowledge attributions. Matt wants subject context and Keith speaker context. Matt uses the principle tying what one knows with what one ought to do to argue for subject context and for pragmatic encroachment into the theory of knowledge. I’m trying to avoid such pragmatic encroachment if it’s possible, and am attempting to do so without resorting to contextualism. The line I was taking in the last comment is that Matt’s principle tying knowledge to action may not imply enough to get pragmatic encroachment, unless Matt converts to contextualism, but my argument depends on a very special decision-theoretic assumptions that Matt might reasonably reject.

Leave a Reply

Your email address will not be published. Required fields are marked *