Do Justified Beliefs Justify Action?

In Can We Do Without Pragmatic Encroachment?, I argued that the various phenomenon that pragmatic epistemological theories were trying to explain were primarily due to the pragmatic nature of belief, not the pragmatic nature of justification. A large part of Fantl & McGrath’s response to this is to argue that a pragmatic theory of belief isn’t sufficient to derive principles like this one, which they take to be central to a pragmatic epistemology.

(JJ) If you are justified in believing that p, then p is warranted enough to justify yoy in φ-ing, for any φ.

This isn’t actually one of the principles they say I can’t derive, but it’s in the ballpark. And it’s relevant because (a) the principles they think I should be able to derive are stronger than (JJ), and (b), (JJ) is false. I think the argument against (JJ) in Can We Do Without Pragmatic Encroachment? is pretty good, but it can be simplified. Here’s a much simpler version. The following is all true of an agent S.

  • She knows that p and q are independent, so her credences in any conjunction formed out of p, ¬p and q, ¬q are products of the credences in the conjuncts.
  • Her credence in p is 0.99, just as the evidence supports.
  • Her credence in q is also 0.99. This is unfortunate, since the rational credence in q given her evidence is 0.01.
  • She has a choice between taking and declining a bet with the following payoff structure.
    • If pq, she wins $100.
    • If p ∧ ¬ q, she wins $1.
    • If ¬ p, she loses $1000.
  • The marginal utility of money is close enough to constant that expected dollar returns correlate more or less precisely with expected utility returns.

As can be easily computed, the expected utility of taking the bet given her credences is positive, it is just over $89. Our agent S takes the bet. She doesn’t compute the expcted utility, but she is sensitive to it. That is, had the expected utility given her credences been close to 0, she would have not acted until she made a computation. But from her perspective this looks like basically a free $100, so she takes it. Happily, this all turns out well, since p is true. But it was a dumb thing to do. The expected utility of taking the bet given her evidence is negative, it is a little under -$8. So she isn’t warranted, given her evidence, in taking the bet.

I also claim the following three things are true of her.

  1. p is not justified enough to warrant her in taking the bet.
  2. She believes p.
  3. This belief is rational.

The argument for 1 is straightforward. She isn’t warranted in taking the bet, so p isn’t sufficiently warranted to justify it. This is despite the fact that p is obviously relevant. Indeed, given p, taking the bet strictly dominates declining it. But still, p doesn’t warrant taking this bet.

The argument for 2 is that she has a very high credence in p, this credence is grounded in the evidence in the right way, and it leads her to act as if p is true, e.g. by taking the bet. It’s true that her credence in p is not 1, and if you think credence 1 is needed for belief, then you won’t like this example. But if you think that, you won’t think there’s much connection between (JJ) and pragmatic conditions in epistemology either. So that’s hardly a position a defender of Fantl and McGrath’s position can hold.

The argument for 3 is that her attitude towards p tracks the evidence perfectly. She is making no mistakes with respect to p. She is making a mistake with respect to q, but not with respect to p. So her attitude towards p, i.e. belief, is rational.

These three points entail that (JJ) is false, since S provides a counterexample. So I don’t think it’s a bad thing that you can’t derive principles like (JJ), or stronger principles, from my theory of belief. The derivation doesn’t work because my theory of belief is true, and those principles are false!

22 Replies to “Do Justified Beliefs Justify Action?”

  1. Hi Brian,

    I’m not sure what you mean when you say that her belief is rational, but I’ll assume that you mean it’s justified (in order to get a counterexample to JJ) which, for the purposes of JJ, means that it’s the attitude she should have of believing, denying, and withholding. With that in mind:

    You say that her attitude toward p, i.e. belief, is rational. I will grant that her credence in p (that is, .99) is rational. What’s the argument that belief in p is rational? After all, if her belief that p is rational, then presumably she could rationally reason like this:

    “p. So, I’ll be throwing at least $1 away by not taking this bet. Of course, if not-p, I’d lose $1000, but, fortunately, p. So I won’t lose money by taking the bet and I will get at least $1. All I care about is money, so it will be best for me to take the bet.”

    When I say that she could rationally reason like this, what I mean is that the conclusion of this reasoning will, unless the case is unusual, end up justified for her (if p is justified for her).
    But if she can rationally reason like this from p to the relevant theoretical conclusions (“It will be best to take the bet,” “I’ll throw money away by not taking it,” “I won’t lose any money by taking the bet”, etc.) then it seems very strange indeed to fail to allow p to be rationally available in practical reasoning as well — unless you’re comfortable segregating all of these reasons from her practical deliberation. The natural way to prevent p from being usable in her practical deliberation is to deny that it is usable in her theoretical deliberation as well. And the natural way to do that is to deny that belief that p is justified for her (even though her credence in p is justified).

  2. The reason that reasoning goes wrong is that she has a messed up other belief, namely q. And when you have other messed up beliefs, even using justified beliefs can take you to bad places. That’s a general feature of ampliative reasoning. (And for the reasons I go over ad nauseum in the last section of the 2005 paper, I don’t think the reasoning in the quoted bit here can be made into deductive reasoning.)

    And I still think the argument I offered in the post is good. S believes that p. Her attitude towards p is justified. (It’s her attitude towards q that’s unjustified.) So she has a justified belief in p. What could be wrong with that argument?

    I think the response you give here slips from doxastic justification to propositional justification. It’s true that she doesn’t have propositional justification for p. But she does, I think, have doxastic justification, because her attitude is one of belief, and her attitude is justified.

  3. I’m not sure why her unjustified belief that q matters. Suppose her credence in q were as the evidence suggests – .01. So as far as her credences go, all is well with the world. We still wouldn’t want to say that the reasoning leads her to a justified conclusion. So it can’t just be messed up q-attitudes that matter.

    Second, if it’s possible to have credence .99 in p without believing that p, then it seems possible to be justified in having credence .99 without being justified in believing. Your argument needs as a premise that belief is the attitude that’s justified. And that premise is question begging.

  4. Hoping I could ask three questions just to get clear on this.

    1. I had thought that Fantl and McGrath’s view was that even if S oughtn’t do various things S did having been motivated to do it by the thought that p, provided that p is justified it isn’t for a lack of justification for believing p that these various things aren’t justified. Whatever else went wrong, S was justified in treating p as a reason even if doing the thing that S did in virtue of having treated p as a reason isn’t justified (e.g., when S believed unreasonably that the thing she did was the thing to do given p).

    2. Quick question about the notion of justification that figures in (JJ). Is the idea that if something is justified (epistemically), it can provide a practical justification? Or, should we allow that there are cases where there’s no practical justification for treating p as a reason but there’s epistemic justification for doing that?

    3. Quick question for Brian on propositional and doxastic justification. I guess it’s somewhat controversial (I think Turri has a paper on this), but isn’t a standard view that propositional justification is necessary for doxastic justification? (Actually, come to think of it, don’t F&M use that assumption in arguing that there can be falsehoods that justify in the dreaded falsehoods can justify section of the book?)

  5. Hi Clayton,

    Not sure what the question is in 1, but I gather that in Brian’s example, it’s not just that the agent isn’t justified in taking the gamble. It’s that her lack of justification for p is standing in the way, because if p were certain for her, she would be justified in taking the gamble.

    On 2 — I don’t think we think of “treating p as a reason” as a separate act that can either be practically or epistemically justified. (Or, rather, if it is a separate act in that way, the fact that it is isn’t really central to the view.) Instead, p can provide justification for acts and beliefs, when p is a reason you have for doing and believing those things. When p provides reasonably truth-related reasons for believing some q, the justification it confers on believing q is presumably epistemic, and when p provides ends-directed reasons for doing something, the justification it confers on doing that thing is presumably practical. But there’s really only one kind of justification there — p makes it the case that you should phi, and whether p does so in an epistemic or practical way is not that important and will probably depend on the relationship between p and phi, as well as, presumably, the nature of phi.

  6. I have no idea from Jeremy’s reply in 3 which premise of the argument fails. Is it that S doesn’t believe that p, or that S‘s attitude towards p is not justified? I don’t think there’s a good answer to this question on his view, but I would like to know what the answer is.

    Of course, if S had the right attitude towards q, she wouldn’t take the bet, and her attitude towards p would not be one of belief. But because she has the wrong attitude towards q, her (perfectly justified) attitude towards p does count as belief. So she has a justified belief that p.

  7. Clayton, I agree that it’s not exactly standard to think that there can be doxastic justification without propositional justification. Indeed, that may not be the best way to put my view. Here are a couple of less controversial ways of putting the view.

    I think mental states are individuated holistically, but evaluated atomistically. So S‘s state counts as a belief because of the way the rest of her mind is. If she had a different credence in q, the very same attitude towards p would not count as a belief. That’s the holism. On the other hand, if we just look at what she thinks about p, then it’s all perfectly in order. So her actual attitude towards p, i.e., belief, is justified.

    Or to put things another way, there are some unexpected scope ambiguities around here. I think that S‘s attitude towards p, i.e., belief, is justified. On the other hand, I don’t think that S is justified in having the attitude of belief towards p. If she were doing everything right, she wouldn’t have that very attitude.

    This might seem like a very fine distinction, but I think we have to make similar in cases where agents are making mistakes in general. Think about the various things we’d say in Michael Smith’s case of the ill-tempered racketball player who just lost. Michael imagines the guy thinking “The best thing to do would be to graciously shake the other guy’s hand and congratulate him for winning. But if I walk over to do that, I’ll probably punch him in the face. So instead I’ll just walk away.” In some sense, walking away is justified; the alternative is awful. But the guy isn’t justified in being a guy who walks away; he should graciously congratulate the winner. Put another way, his actual action of walking away is justified-for-him, although the action isn’t objectively justified. (That’s the parallel to S‘s belief being doxastically but not propositionally justified.)

    I think S is in a similar position. Her actual attitude towards p, i.e., belief, is justified. But she shouldn’t be in the position she’s in, so she shouldn’t in some sense believe p.

  8. Jeremy,

    As in the response to Clayton, I think there’s a sense in which S shouldn’t believe that p. What I don’t see is how that makes S‘s belief unjustified. She shouldn’t believe q, and if she didn’t, her actual (and justified) attitude towards p would not be belief. But she does believe q, and, more importantly, the only mistake she’s making is a q-related mistake. I don’t see how to motivate the thought that her attitude towards p is unjustified.

  9. She bears two different relations to p — belief and credence .99. I don’t see why the evaluation of her token state wrt p should be fixed by the evaluation of her instantiation of the relation of credence .99. Here’s a reason not to evaluate her token state that way: it means we have to allow for the possibility that S’s belief that p is justified even though S shouldn’t believe that p. Here’s another reason not to: if her belief that p is justified then her belief that she’ll in effect throw away at least a dollar by refusing the bet is also justified, which means she is justified in taking the bet.

    In any case, I now fail to see how this is a counterexample to JJ. If you’re granting that she shouldn’t believe that p, then you’re granting that the antecedent of JJ is false, which means it’s no longer a case in which the antecedent is true and the consequent is false.

  10. I don’t see why there’s any reason at all to believe this:

    if her belief that p is justified then her belief that she’ll in effect throw away at least a dollar by refusing the bet is also justified, which means she is justified in taking the bet.

    The last clause simply doesn’t follow from the first two. After all, it doesn’t follow from the fact that I’d make money by placing a bet that I should take it. (Every roulette wheel has a winning number, which I would make money by betting on, but I shouldn’t bet on any roulette wheels.) So “make money” → “good bet” isn’t always true. So without further reason, I think Justified in believing (make money) → Justified in believing (good bet) isn’t always true. And this seems to be a counterexample to it.

    And the last paragraph is odd too. Just like in the racketball example, what she should do and what’s she justified in doing come apart. She’s justified in believing p. She shouldn’t be the kind of person for whom that’s true, but she is. Indeed, she shouldn’t be the kind of person who believes p in this circumstance, but she is. Similarly, the racketball player is justified in walking away, though he shouldn’t be the kind of person who walks away.

    I might have been a little careless on this earlier. In one sense, the most important sense I think, S should believe that p. She should have the attitude she actually has, and that attitude is belief. (For the reasons I went over in the earlier paper, I think it’s a very bad idea to say the belief and credence are separate attitudes. They are separate ways of describing the same attitude.) On the other hand, she shouldn’t be the kind of person who is best described as believing that p. So in that sense she shouldn’t believe that p. But that’s a misleading sense, it’s the same sense as in which the racketball player should walk over to his opponent.

  11. Brian,
    Jeremy points out that if she was justified in p, she could reason this way:

    “p. So, I’ll be throwing at least $1 away by not taking this bet. Of course, if not-p, I’d lose $1000, but, fortunately, p. So I won’t lose money by taking the bet and I will get at least $1. All I care about is money, so it will be best for me to take the bet.”

    If I understand you right, Brian, you say this reasoning “goes wrong is that she has a messed up other belief, namely q.”

    But the reasoning Jeremy gives doesn’t involve q as a premise at all, or not-q. I don’t see why one’s attitude toward q matters at all to the quality of this reasoning.

    Now, if that reasoning is good, then it looks like (1) below will be true:

    (1) If she is justified in believing p, then she is justified in believing that taking the bet will have better results than not taking it (i.e., justified in believing that the actual utility of taking it is higher than that of not taking it).

    Finally, Jeremy and I want to argue that (2) is true:

    (2) If she is justified in believing that taking the bet will have better results than not taking it, then she has an excellent and undefeated reason to take the bet rather than to refuse it, and therefore she is justified in taking the bet.

    Since we think she is not justified in taking the bet, through a couple of modus tollens we conclude that she is not justified in believing p.

    Where precisely do you think this goes wrong? At (1)? At (2)? And do you grant that whether her belief or credence in q is justified or not is irrelevant? If not, how is it relevant, given that the reasoning Jeremy presents does not involve q at all?

  12. I think the reasoning is defeated by q, and I think it isn’t that uncommon for beliefs to be defeated by things that look somewhat external. Consider this case.

    I’ve seen lots of blue swans, and no swans that aren’t blue. I’m told by Slartibartfast that there is a green swan in the river outside my window. But I think Slartibartfast is a silly name, and I shouldn’t listen to people with silly names. I then reason to myself as follows. “All the swans I’ve seen so far, and I’ve seen swans in a lot of different places, have been blue. So if I see a swan in the river today, it will be blue.”

    Now I haven’t used Slartibartfast’s testimony anywhere in that reasoning, but the reasoning is clearly bad. Indeed, it’s defeated by Slartibartfast’s testimony. So I don’t think that the fact that some proposition isn’t appealed to in reasoning, even reasoning that looks on the face of it generally good reasoning, means that proposition can’t be a defeater for the reasoning.

    That’s to say, (2) is false. S has at most a defeated reason to take the bet. I’m not sure whether she has any reason to take the bet – I suspect in these cases only knowledge, not justification, suffices for providing reasons. But in any case, she certainly doesn’t have an undefeated reason to take the bet.

  13. We note in the book that we are committed to the claim that the property of being a belief that p and the property of being credence .99 in p are distinct properties (and the same goes for any other specific or chunky credence). I’m not sure if you’re denying this. I’m happy to grant that the token state the agent is in when she has .99 credence in p is the same token state that the agent is in when she believes that p, in this case. But maybe you don’t want the property of having a belief to be distinct from the property of being in some specific credal state.

    But the reason to believe the passage you quote is the argument I gave in the opening comment. The roulette example is not to the point, because that is not an example in which the potential gambler is justified in believing that she will win money by betting on the roulette wheel. But in your example, if she is justified in believing p, then why wouldn’t she be justified in believing that the worst she could do by betting would be to get one dollar? After all, even if q is false, she’ll get a dollar, as long as p is true. And, she justifiedly believes, p IS true.

    Do you want to say that even if she justifiedly believes p, she can’t use it to reason to theoretical conclusions she knows figure in true conditionals with p as an antecedent? For, in your example, she surely knows this (and not just barely): if p, then the actual utility of taking the bet will be higher than the actual utility of turning it down. And, in this case, then, she can justifiedly believe the consequent.

  14. In response to your response to Matt:

    I don’t see how q defeats the reasoning, here. Whether or not q is true, if p is true, the actual utility of taking the bet is higher than the actual utility of refusing it. I take it that you are precluded from saying that q defeats the support for p, on pain of saying that belief in p is no longer justified. So, q must defeat the inference from p to the claim that the actual utility of taking the bet is higher than the actual utility of refusing it. Why would it?

  15. Right, she justifiably believes p, she justifiably believes she’ll make money, but she isn’t justified in moving from that to “I should place the bet”. I don’t see what kind of reasoning could justify that last step. It isn’t deductive reasoning, since the inference is invalid. And if it’s any kind of inductive reasoning, then any kind of irrational belief in the vicinity has the potential to be a defeater. (That’s the point of the swan example in my previous comment.) Since she has an irrational belief that’s pretty relevant here, namely q, I think the inference is defeated. I haven’t seen a reason to think that’s wrong.

  16. Oops, the last two comments crossed. I don’t think q defeats the inference to “Taking the bet maximises actual utility”. But I don’t think a justified belief that taking the bet maximises actual utility suffices for action. Especially since in this case, the bet doesn’t maximise evidential expected utility. If she knew that taking the bet maximises actual expected utility, things might be different.

  17. Brian, if you accept 1, you’re accepting the reasoning Jeremy mentions (remember ‘best’ is understood in terms of actual utility). Right? So you’re ok with that?

    If you could give a story like that the Slartibarfast one either for the reasoning Jeremy mentioned or for the piece of practical reasoning from “taking the bet will have better results than not” to the decision to take the bet, I’d like to see it. In the Slartibarfast case, there is something the person is justified in believing which, when added to the premises, clearly undermines the conclusion. But this structure isn’t found in the case at issue, as far as I can see. I understand you aren’t saying the cases are necessarily parallel. We we need a reason to think this broad kind of defeat is present here.

    It might be a while before I check in again, probably tonight.

  18. Thanks, Brian. Some further clarificatory questions:

    1) Do you think that p is a reason she has to take the bet, but that the reason is defeated by a countervailing reason? Or that, even though she justifiedly believes that p, p isn’t a reason she has to take the bet?

    2) When she justifiedly believes that taking the bet will get her more money, is that a reason she has to take the bet — a reason that is defeated by some other countervailing reasons?? Or do you think that the fact that taking the bet will get her more money is not a reason she has for taking the bet, even though she justifiedly believes it?

  19. I think I’m pretty much committed to (1). After all, I think S is justified in believing p, I think p plus some stuff that’s not in question obviously entails that taking the bet maximises actual utility, and I accept various closure principles that are strong enough to entail that S is justified in believing that taking the bet will maximise actual utility.

    But from this point on, I think it’s exactly like the Slartibartfast case. S has a belief that usually, but not always, supports a particular conclusion. She has extra evidence that this is one of the unusual cases where the conclusion is not supported. (In one case, testimony of a green swan, in the other case, direct evidence that the expected utility of the bet is negative.) She ignores this and goes ahead and draws the conclusion. That’s bad, and it’s bad for the same reason in each case, that she’s ignoring defeaters of non-deductive inferences.

    I’m not sure what to say about the reasons. I’m sort of tempted to say it is some reason, but one that’s defeated by her other beliefs. But I don’t think much turns on it.

    I think I should be clearer about how important the token identity is to my way of thinking about this. I think there’s one bit of S‘s mind that we theorists can represent in one of two ways, either as a belief that p, or as a credence of 0.99 that p. Since it’s the one thing in both cases, it can’t be subject to different normative evaluations, on pain of violating Leibniz’s Law. I assume Matt and Jeremy have a very different model of epistemology, perhaps that we evaluate minds (as a whole) in virtue of their instantiating or not instantiating various properties. So S is good in virtue of having the has credence 0.99 in p property, and bad in virtue of having the believes p property. This seems to me to get the evaluations in at the wrong level, but it would help out of my counterexample.

  20. It’s kind of crucial to know exactly what the defeating reason is in this case. It can’t really be q, because if q is true, that contributes, if anything, to a reason there is to take the bet. It must be something like, “But q is really quite likely false.” Or, maybe, even, not-q.

    But how would the reasoning go, here. “p. So I should take the bet. But, not-q. So… I still should take the bet. Because if p and not-q, I should take the bet.”

    That can’t be it.

    Maybe you’re looking for something like this:

    “p. So I should take the bet. But, not-q. So, for me to take the bet, p had really better be true. But, you know, there is a chance it’s not. And it’s a significant chance. So, I guess I shouldn’t take it.”

    (Here I’m not suggesting that she actually goes through this reasoning — after all, she has a really high credence in q. But this is the model of the reasoning that would have to be good if p is somehow defeated as a reason to take the bet.)

    If this reasoning is good, it allows p — a reason to take the bet — to be defeated by the chance that not-p (a reason to refuse the bet). That possibility is problematic for two reasons. First, such reasoning sounds absurd (well, in normal situations, of which yours is one). It sounds absurd to weigh facts against the chance they’re wrong. “On the one hand, there’s a chance p is false. On the other hand, p is true. Hmmm… which is more important — the chance it’s false or the fact it’s true?” That this sounds absurd suggests that we never actually do have both reasons at once. Second, if we could have both reasons at once, it seems to me that the facts should beat out the chances because, after all, the fact entails that the chance won’t obtain and, furthermore, we care about what will actually happen. When what will actually happen is a reason you HAVE, it should beat out what might not happen, when that’s a reason you have.

    Okay. We can get into an argument about this, but if I might indulge in some Fantl/McGrath scholarship, it seems to me that your fundamental target is not actually JJ, but rather what we call Safe Reasons. The only step in the argument for JJ that actually includes mention of justified belief is a step you agree with — that justified belief is a reason for belief. Safe Reasons, on the other hand, says that if p is a reason you have to phi then p is warranted enough to justify you in phi-ing. So, if p is a reason you have to phi, and p fails to justify you in phi-ing, it can’t be epistemic failures wrt p that are standing in the way. In particular, in normal cases, p can’t be defeated by chances associated with its falsehood.

    In the Slarti case, your reason for believing that there will be no non-blue swans is not defeated by chances associated with its falsehood. Even if it were certain for you, it would still be defeated by Slarti’s testimony.

    Again, I’m happy to bicker over Safe Reasons if you like, but I think I can at least say that it is by using that principle that we would resist your counterexample.

  21. I think that’s right – if I think p is a reason in this case, I’m committed to denying Safe Reasons.

    But it isn’t the only move available here. I could simply deny that justification suffices for reasonhood. That’s what Stanley and Hawthorne do, for instance. And there’s a reason for that denial. I think cases like S‘s belief that p are very odd kinds of JTB without knowledge. (So odd that I think it’s misleading to call them Gettier cases.) The argument that justified belief suffices for reasonhood in KIAUW is, if I’m reading it right, that the differences between J and K aren’t relevant to reasonhood. If I’m right, there’s a new kind of difference between J and K, one that the various kinds of subtraction arguments in the book don’t account for.

Leave a Reply