Necessary Epistemic Principles

While at Brown, a quite interesting issue came up a couple of times, both during one of the sessions and also in converstation, that I thought would be good to pursue here. A common assumption is that what divides epistemologists with my sort of inclinations in the theory of justification from those such as Goldman, Nozick, Sosa, Greco and the like has something to do with the necessity, and perhaps a priority, of epistemic principles. While it is true that internalists such as Chisholm view epistemic principles in this way, I have never quite been able to see why this issue would be seen as a dividing issue between internalists and externalists.

Here’s why I’m perplexed by the assumption.

Every theory of justification will generate some necessarily true epistemic principles, so long as the theory itself is taken in the usual philosophically robust sense, i.e., is taken to be necessarily true if true at all. This point holds for internalist and externalist theories alike. But the idea is supposed to be that all the principles of a certain kind of theory have to be necessary. I’m inclined to think there isn’t a prayer for an argument to this conclusion unless we’re thinking of access internalism here. In such a case, the principles that guide belief formation would need to be a priori, and perhaps there is some way to get from a prioricity to necessity in this case (even though we can’t in general draw such a conclusion).

Take, for example, classic bayesian coherentism, where probabilistic coherence is necessary for justification and updating occurs by conditionalization. Presumably, conditionalization is regarded as a necessary truth, but not any specific principle that ties a particular new piece of knowledge (to speak loosely) to updated degrees of belief (i.e., more specific principles that play an appropriate role in explaining the state transitions in question).

But maybe there is some argument I haven’t thought of which ties a denial of externalism to the idea that epistemic principles have to be necessarily true if true at all.


Necessary Epistemic Principles — 34 Comments

  1. I agree that all parties will agree that the correct epistemic principles are necessarily true. Maybe something along the following lines would be more fruitful.

    Suppose an experience as of a hand is a pro tanto good reason (or: prima facie evidence) for me to believe that there is a hand. Does it follow that, necessarily, for all S, an experience as of a hand is a pro tanto good reason for S to believe there is a hand?

    I think that’s an interesting question. I don’t know if it’ll neatly divide internalists and externalists, but it seems to hold out at least some promise. In my judgment, the following epistemologists would answer “yes,” though perhaps not for the same reasons:

    Conee and Feldman

    I can’t think of any prominent externalists who would answer “yes.” Simple reliabilists would certainly answer “no.” I’m thinking that Plantinga would say “no” because he’d want to tie good reasons and evidence to design plans, and not all design plans are going to include the relevant specification. I don’t see a Nozickian tracking theorist answering “yes” unless he’s willing to completely divorce questions about evidence from his treatment of knowledge. More sophisticated versions of reliabilism say “yes and no.” Sosa and Comesana, for example, have the resources to say this, as do some time-slices of Goldman.

  2. John, very nice here. What about Lehrer, for another example? I’m inclined to think he’d say no, because I know he’s always thought that Chisholmian principles weren’t necessarily true. I think Bayesian coherentists would agree with Lehrer here.

    This makes me wonder why you think BonJour would say yes. I would think that the answer for him (I’m thinking of the earlier BonJour, not the later one) is going to depend on which beliefs are spontaneous, and that would be contingent. Or were you thinking of the later BonJour?

  3. It is difficult to see why anyone–including bayesians–would think that probabilistic coherence is *necessary* for justification. Inductive principles in general seem rather obviously dependent on contingent features of the actual world. Certainly there is a world in which the consistent observation of event e in C (say flipping “heads” in normal circumstances) is evidence that event e won’t happen next time. That is, there are worlds in which the gambler’s fallacy is not a fallacy at all (perhaps ours will turn out to be one). Similarly for other inductive principles.
    The point seems to generalize. Given the possiblity of wide variation in our epistemic relation to the world or to evidence, it is hard for me to see why epistemic principles ought to be necessarily true (if true at all). Perhaps the necessity involved is less broad that what I have in mind.

  4. Mike, I’m inclined to agree with you about both points. The necessary truths will be at a very abstract level, where the entire theory itself is formulated. Your gambler example sounds a bit like an externalist example, but I don’t think it needs to be: embedded in the right background system of information, and you’ll still be able to get this result.

  5. Jon,

    Right, I forgot about Lehrer, so not all parties will agree. But his denial is bound up with his argument against the supervenience of normative epistemic properties, which I don’t understand.

    Yes, I was thinking later BonJour.

    I’m curious whether there are any (actual) externalists who’ll simply say “yes.” If there aren’t any, then I think the question I posed at least points in the direction of one factor underlying the internalist/externalist divide.

  6. John, I think it depends if you want a remotely plausible externalism! Here’s a rather implausible one. Suppose justification is defined in terms of probabilistic relations, and one hold a logical theory of probability. Then the probabilistic connections will all be logically necessary if true, and hence the justificatory relations will all be necessary if true as well.

  7. Jon,

    Yeah, I was trying to stick to remotely plausible ones- -or at least ones actually on offer in the literature. šŸ˜‰ (Pretty soon I won’t be able to say that, since this blog will count as “part of the literature.” Or maybe it already does?)

    I probably should have added a caveat to the way I put the question, since at least some people whom I said would vote yes do add the caveat. And that would be to restrict the range of S to subjects in exactly similar mental conditions as the one I’m presently in, whether it be occurently conscious mental conditions, total mental conditions, or somewhere in between. This gets us into all kinds of controversy about mental content, which might muddy the waters even more. If the requirements for same content are stringent enough, then we might well get some externalists answering “yes.”

  8. I’m with Mike. Take this principle: if it looks like a duck, walks like a duck, and quacks like a duck, it’s probably a duck. The truth of this is entirely a posteriori; it depends on how many convincing duck-fakes are around, for one thing. In fact, if we know on other grounds that all the normal-looking -walking and -quacking ducks have been eliminated, and only grossly deformed ducks remain, then looking/walking/quacking like a (typical) duck will be evidence *against* being a duck. So if there are any necessarily true or a priori epistemic principles, they have to be at a very high level of abstraction.

  9. Juan,
    Thanks for the link to your paper. I’m wondering about a short argument in section II. It goes this way:
    “Could someone doubt that there are support facts so conceived? Hardly so. Surely, if p were not a good reason to believe that q, then Sallyā€™s being justified in
    believing that p would not justify her in believing that q.”

    Can that be right? Here’s a small counterexample:
    Let p = it is raining outside.
    Let q = S has a justified belief.
    If S is justified in believing that p (= it is raining outside) then S is justified in believing q (= that she has a justified belief). But the fact that it is raining outside is not a good reason for believing that S has a justified belief. So p is not a good reason to believe q but the fact that S believes that p does justify S in believing that q. Contrary to your argument above, I think.

  10. Mike,

    Thanks for the comment. I’m not at all sure that if S is justified in believing that it is raining outside then S is justified in believing that she has a justified belief. Maybe S doesn’t have the concept of belief, or maybe S is an epistemologist with a false (but perhaps justified) theory about what it takes for a belief to be justified.

  11. Juan,

    In the example I don’t assume that S knows that she is justified in believing she has a justified belief. I say only that she is justified in so believing. On the other hand, if she is justified in believing p only if she knows that she is justified in believing p then S knows that she is justified in believing that it is raining outside. It is hard to see how that evidence wouldn’t justify S in believing that she has a justified belief. That’s the counterexample. It does raise closure issues, but that doesn’t seem to be a problem you’re discussing.

  12. Mike,

    What I said I find highly dubious is that S is justified in believing that she has a justified belief in virtue of being justified in believing that it is raining outside (not having the concept of belief, or having a mistaken conception of justification precludes this, I think). I also think, of course, that she doesn’t know that she is so justified, but that is besides the point. And no, I don’t think that S is justified in believing that p only if she knows that she is so justified.

  13. I think we can find out quickly what I’m missing.
    1. S is justified in believing that it is raining outside.
    2. :. S is justified in believing that she has a justified belief.

    You say that this inference is probably mistaken because (for instance) S might not have the concept of belief. Suppose S doesn’t have the concept of belief and also doesn’t have the concept of canine.

    1′. S is justified in believing there is a dog in front of her.
    2′. :. S is justified in believing there is a canine in front of her.

    The fact that S does not have the concept of canine does not–not as far as I can see–preclude the inference to (2′). She is justified in believing there is a canine in front of her, though she doesn’t know what canines are and could not draw the inference herself. If you’ve got good evidence that there is a dog there, then (whether you know it or not, or have the right concepts or not) you’ve got good evidence that there is a canine there. This looks to me perfectly consistent with the opacity of knowledge contexts.

  14. But there is a relevant difference between your two cases, Mike. In the second one, the substitution of ‘dog’ for ‘canine’ occurs inside the belief-context, whereas in the first one you want to have the subject conclude something on the basis of his having a given attitude. I might agree with you that if a subject is justified in believing outside then she is justified in believing that it precipitates, but that is not a counterexample to my claim.

  15. The cases are analogous with respect to probabilistic closure. Suppose (1) is true and so has probability 1. So it is true that S is justified in believing that it is raining outside. But certainly, ‘S is justified in believing that it is raining outside’ entails ‘S has a justified belief’. So the probability that S has a justified belief also has probability 1. But if it follows from (1) that the probability is 1 that I have a justified belief, then on the basis of (1), I am justified in believing that I have a justified belief.

  16. I’m starting to loose track of the issues, here. I agree that 1. entails that S has a justified belief. How does that support the idea that S is justified in believing that she has a justified belief, if it is open that S doesn’t even believe 1 (let alone that 1 entails that S has a justified belief)?

  17. Juan, I’ve just read the paper rather quickly, but it seems to me (and maybe you’d agree) that it might provide an answer to Jon’s original question: Why would internalists be especially likely to see epistemic principles as necessary or a priori?

    The general version: Suppose you define internalism according to principle M, that mental duplicates must be epistemic duplicates. (That’s the version I’d subscribe to–except I’d say, if you’ve had exactly the same experiences you must be epistemically the same; and that might include past experiences, and experiences might not be internal. But my views on that are by the by.)

    Now, you should frame specific epistemic principles in terms of the entirety of the subject’s mental states–so that it’s “Anyone who has this total mental state should believe q.” (Otherwise you’ll miss out on defeaters–of course the sensible thing to do is frame general defeasible principles.) Then, if mental duplicates are epistemic duplicates, such a principle has to be necessary, because every counterpart in every world with that total mental state ought to believe that q. (And if we substitute ‘experiences’ for mental states, then experience can’t make a difference to the principle either, which gets you some way to aprioricity.)

    The version that’s more specific to your arguments: Let’s say you see internalism as meaning, not that justification requires actually attaining the mental state in which you are aware of the reason for your belief, but rather being in a position to reach that state. So the support fact (‘Experience e supports belief p’) isn’t a mental state that the believer must have in order to be justified in believing p on e, but it is a fact that the believer must be in a position to grasp in order to be justified. (And there’s another fact that supports belief in that fact—but if we don’t require actual grasp of the fact, merely being in a position to grasp it, then there needn’t be a vicious regress; the believer could be in a position to grasp each fact after grasping the one before.) The only way to sustain internalism then will be if the support fact either depends on mental states—which you argue against convincingly—or is graspable irrespective of your mental or external states, which is to say graspable a priori.

    (As for the argument that support facts may not be necessary, I’d rely on the totality of experience/mental states line again. If Steve is in an environment where induction fails, but all his experiences have been exactly like those he’d have in an environment where it succeeds, then he is justified in believing that all swans are white. So support facts would actually be “Experience-totality E supports belief p,” and those are more plausibly necessary—for internalists, anyway.)

    OK, I’m not convinced that the latter part really responds to what you’re getting at, but it’s pretty late…

  18. Hi, Matt,

    This isn’t quite what I’m looking for as a defense of necessity, since in this sense, reliabilists think there are necessarily true epistemic principles as well. They think, for example, that once you fix the process and the environment, then you’ll fix the justificatory facts.

    So, where does contingency come in for them. At the level of that which explains the holding of the belief in question: what we might characterize in terms of the basis of the belief. That means that if we have an account of the necessity of epistemic principles, it ought to be the necessity of those principles involved in the explanation of the doxastic justification of a belief, not just the propositional justification of a belief. In other words, we want to know how basing is characterized, and what principles linking epistemically supporting factors with belief are involved in the basing relation.

    These principles will not be holistic ones citing in their antecedents everything about the mental facts of the individual in question, on pain of having to hold that doxastically justified belief is based on such a totality. That would render every version of internalism susceptible to Pollock’s objection to holistic coherentism. But once one formulates the principles at work in basing, I don’t see how those principles are going to be necessarily true.

    The best case for necessity here is to think that the relevant principles will have both positive support cited as well lack of grounds for doubt in their antecedents. I have two worries about such a proposal. The first one I posted about earlier: I think the default principles operative in belief formation do not have grounds for doubt clauses in them. One way to put the point is that we learn to be wary rather than being that way naturally. The second worry is that context affects what is learned from experience. Of course, that is the coherentist inclination in me, but I think it is instructive since I think it puts a burden on the internalist to not be a coherentist if desiring to affirm the necessity of the principles in question.

  19. Matt,

    I am sympathetic to your parenthetical remark regarding the necessity of support facts—and in the paper I say something to the effect that that is the best line for someone wishing to defend the necessity of support facts.


    I don’t think that reliabilists come out as necessitarians in the sense that Matt and I (and John earlier) are thinking of: they would deny the supervenience of epistemic status on mental states because they think that, say, whether an experience as of a hand is a pro tanto reason to believe that there is a hand depends on whether that experience is reliably connected to the truth of that belief. Indexical reliabilists, as John points out, are a different matter.

  20. Juan, yes, you’re obviously right that reliabilists aren’t necessitarians in the sense you and Matt are talking about. That wasn’t my point, though–I was only wishing to point out that the necessity/contingency issue won’t divide externalists and internalists if we’re at such an abstract level, since reliabilists will have necessary epistemic principles at this level of abstraction as well.

  21. One last try. Maybe this is clearer. You seem prepared to say that (2) follows from (1) (it’s your example),
    1. S is justified in believing that it is raining outside
    2. S is justified in believing that it is precipitating.
    And you seem prepared to say this on the assumption that S has no idea that (1) is true or has any idea what precipitation is or for that matter what a belief is. But (1) is no stronger evidence for (2) than it is for (2′).
    2′. S has a justified belief that (Ep)S believes p.
    I think I understand your reservations about (2′) and its ilk. But (2′) does seem to follow from (1) for this reason. If you put the probability of ‘(Ep)S believes p’ at something close to 0, then the probability of (1) is close to 0. So it follows from the fact that S simply HAS the justified belief in (1)–setting aside altogether whether S knows she has it or suspects she has it and setting aside the array of concepts S has (or fails to have)–that the probability of ‘(Ep)S believes p’ is quite high. All she has to do is HAVE the justified belief in (1) and the probability of ‘(Ep)S believes p’ is high. But she has the belief in (1), by hypothesis. What more does she need to be justified in believing (Ep)S believes p? Of course she might have no idea that she is justified in believing that (Ep)S believes p, but that’s not relevant here.

  22. Mike,

    Here’s the difference that I see between (2) and (2′): S has a justified belief with the content that it is raining, and so she is justified in believing that it is precipitating (because that it is raining is a good reason to think that it is precipitating); S has no belief with the content that S is justified in believing that it is raining, and hence she is not (in vitue of being justified in believing that it is raining) justified in believing that she has a justified belief. I agree that that one is justified in believing that it is raining is a good reason to think that one has justified beliefs, and so, if one were to have a justified belief with the former content one would be justified in believing the latter. But (again) S doesn’t have a belief with the former content.

  23. Juan, yes, I see that. I know that’s your reservation about (2′). *Having* the belief in (1), you want to say, is not enough to license an inference to (2′). That I have the belief in (1) must also be an object of belief and so on.
    I’m suggesting that ‘having the belief’ is enough to justify me in believing that (Ep)I believe p because *having the belief* just is *having strong evidence* for the claim that (Ep)I believe p. I think you want to say that it is evidence that in the relevant sense I do not have (since I don’t have the right sort of belief about it). But then there is in my perceptual field a salient figure of a phone. I’d say that I’ve been justified in believing there’s a phone nearby, but I simply haven’t, (until just now) realized that I possess the evidence for that claim. Similarly I’d say I have the evidence for (2′)–I have the belief in (1)–I just don’t realize I have it. Anyway, it’s pretty clear that we disagree right about here.

  24. Mike, I agree that that’s were we disagree. (I would resist the analogy with the phone, though—depending on the details, I’d say that you are right that you are justified in believing that there’s a phone nearby even if you haven’t been paying attention to it (although there are delicate issues here), but I still think that you are not justified in beliving (2′) in virtue of believing that it is raining.) And by the way, it is not (just) that that’s what I’m committed to saying: I find it obvious that you are not justified in believing that you have a justified belief simply in virtue of having a justified belief.

  25. I thought I would take a shot at explaining why internalists are committed to necessary/a priori epistemic principles. Here I am thinking of EPs (with Chisholm) as principles stating conditions sufficient for a belief having some epistemic status, where the antecedent specifies states of the believer, such as the believer having some evidence of a sort, propositional or otherwise.

    So here is the explanation. First, internalists think that the facts that contribute to justification are in some important sense internal. Second, since the fact that evidence E is good evidence for p contributes to justification, internalists of different stripes tend to endorse something like Fumerton’s Principle of Inferential Justification:

    (PIJ) To be justified in believing one proposition P on the basis of another proposition E, one must be (1) justified in believing E and (2) justified in believing that E makes probable P.

    Hume teaches us that “probable” can’t mean objectively probable in PIJ, on pains of skepticism about empirical knowledge. Fumerton, for example, sees this clearly. Therefore internalists opt for a sense of “probable” that is epistemic–something close to “epistemically justified on the evidence,” where this does not imply anything like de facto reliability or objective probability. In other words, “E makes probable P” is a necessary truth about concepts or meanings or some such thing, rather than a contingent truth about the world. In other words, propositions of the form “E makes probable p” are necessary truths if true at all. But “E makes probable p” sure looks likes the form of an epistemic principle.

    Externalists are not motivated to accept any such thing, since they are not motivated to accept either PIJ or the relevant notion of epistemic probability.

  26. “In other words, “E makes probable P” is a necessary truth about concepts or meanings or some such thing, rather than a contingent truth about the world.”

    But don’t we want to say that knowing that E = The coin was flipped two hundred times and came up “heads” two hundred times, justifies me (given certain qualifications) in believing P = the next time the coin is flipped it will come up “heads” Isn’t it true in this case that E makes probable P? I think so. But whatever one’s conception of probability, it is hard to see this claim as a necessary truth.

  27. Hi, John, I think you are right about foundationalist internalists, but not about coherentists. Remember that one of Lehrer’s famous dicta is that he always thought Chisholm’s principles were true, but just not necessarily true.

    Of course, even for coherentists, there is some necessary truth in the neighborhood: that it is necessary that E justifies p relative to a particular system of information (or something along these lines). But every theory will find necessary truths somewhere.

    The best test, I think, is to look at the principles that are operative regarding proper basing. Here contingency reigns for both internalists and externalists. This is the proper place to look for differences between internalism and externalism, since externalists are talking about what is most often called doxastic justification (and what Feldman calls well-foundedness). So if the necessity/contingency point were central, it should appear here. But it doesn’t.

  28. Mike,

    I think that at least some people who read “E makes probable p” as something like “E epistemically justifies p” would take this as a necessary truth. I am thinking of people like Chisholm, BonJour (early and late), Feldman and Conee, Fumerton, Moser? all who think that the relevant notion of epistemic probability/epistemic justification gives us either analytic or synthetic apriori truths about evidential support relations.

  29. Jon,

    I would think that the difference would be found in views about evidential support relations. Externalists tend to think that whether some E supports some p is a contingent truth, depending on facts about cognition and the world, whereas internlists tend to think it is a necessary truth. Here the issue seems to be essentially about content relations, abstracting away from whether anyone in fact believes p on the basis of E.

  30. John, I think that is wrong. For everybody, there’s some necessary truths about epistemic support, but it’s just not true that the justification relation between propositions is always thought of as necessary by internalists. I don’t think they need to be, and neither does Lehrer. What, for me, is necessarily true is that justification depends on the totality of your evidence, and that evidence is thought along internalist lines of some sort. If you specify the total evidential situation and the proposition in question, then there will be a necessary truth about the justificatory status of that proposition relative to that total evidential situation. But there will also be a necessary truth in the neighborhood for reliabilists. The claimed necessary truths will be different, of course: the reliabilist will include external conditions in the claim, and internalists won’t. But the truths that we use to explain the justificatory status of a given belief will go beyond these necessary truths, and need not be necessary truths themselves, even for internalists. For some, such as Chisholm (and I think for Feldman and Conee as well, though I remember a conversation with Rich in which he speculated that perhaps the only necessary truth in the area was one that connects justification with total evidence, leaving contingent whether any particular bit of information counts as evidence in a given case), all principles are thought to be necessary; but for others, they are not. The analogy with reliabilism is straightforward, I think: the contingent principles of reliabilists are ones having to do with doxastic justification, and hence involve proper basing; the relevant principles for internalists simply can’t be necessary truths, as argued above.

  31. John,

    I must be missing something in your comments. It looks like you’re asserting the following conditional (on behalf of some internalists):

    If (E makes probable p) then [](E makes probable p)

    Now suppose the total evidence I possess for p (= the next ball from the urn will be red) is E (= the last 1000 balls from the urn have been red). E makes probable p only in worlds where reverse induction is fallacious. So it looks like you are saying that either (i) reverse induction is fallacious in every world or (ii) E doesn’t make probable p. If (i) then the “problem of induction” is solved a priori–there is no such problem, and if (ii) then some widely approved inductive reasoning is mistaken. I don’t think you want to say either of these. So how have I misread your comments?

  32. Mike and Jon,

    Sorry for the long delay. I am not a very responsible blogger.

    Anyway, it still seems to me that an important difference between internalists and externalists (maybe the most important) is their attitudes toward epistemic principles of the kind Chisholm used to defend– i.e. principles specifying relations between states of the believer and epistemic status of beliefs. They would have a form such as “If S is appeared to F-ly andS has no grounds for doubting. . . then S is justified in believing that some x is F. Externalists think these are contingent. Internalists tend to think they are synthetic a priori, although some think they are analytic. Coherentists such as Lehrer think that we need to specify a very broad state, much broader than Chisholm ever did, but that does not seem to me to be important here. What is important is whether the relations between cogntive states (amounting to S’s evidence or grounds) and justified belief are a priori or contingent.

    If we interpret “probable” as “epistemically justified”, then corresponding to such EPs will be propositions of the form “E makes probable p”. Again, coherentists lik Lehrer will think that a lot more goes into E then Chislolm did, but both think that, when specified correctly, the relation is a priori. Externalists will reject this. Whether some E justifies some p, they will say, depends on such things as whether E is a reliable indication of p, and that varies in different worlds.

    So maybe another way to put the difference is this– internalists think that you get necessary truths at the level of Chisholm-type EPs specifying relations between grounds and beliefs. Externalists think you need to bring in additional facts (for example about relaibility) before you get to necessary truths.

Leave a Reply

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