Access Internalism and JJ principles

Ralph’s interesting post and penetrating discussion raised to my mind the central question of the relationship between access internalism and various principles about the connection between first-order and higher-order justification. The simplest are, where ‘p’ is a proposition and ‘J’ is the justification operator which can be read “it is justified that” (with the understanding that it is the same person with the same total epistemic condition between antecedent and consequent), these:
1. Jp entails JJp
2. JJp entails Jp.

Here I have no special interest in 2, and don’t think anyone, even access internalists, ought to endorse it. Perhaps a defeasible connection should be endorsed (if JJP and no external defeaters with respect to JP, then JP), but the unqualified principle strikes me as obviously false (for the same reasons that infallibilism in general is false). But principle 1 is more plausible to me (when the kind of justification is the kind figuring in an account of the nature of knowledge, i.e., when ungettiered and combined with true belief yields knowledge), and the relationship between it and access internalism is the topic here. In short, access internalists, I wish to maintain, move quickly to operationalizing principle 1, in a way that leads to problems for their view that do not threaten principle 1 itself. More below the fold…

An access internalist rendering of principle 1 says: if you are justified in believing p, then if you reflect on the question of whether you are justified in believing p, it will be really obvious to you that you are. This rendering has all of the faults of conditional analyses of anything, being threatened by Shope’s conditional fallacy. For example, some people, when they reflect, can’t get past their own inability to come to a conclusion on the basis of reflection. Others mistrust their reflective capacities, and so doubt what seems to be the right conclusion. And others still reflect in such a way that additional evidence is created by reflection that didn’t exist prior to the reflection. In all these ways, an access internalist construal of principle 1 has difficulties to solve.

Note, however, that principle 1 doesn’t require access internalism. In fact, access internalism is merely what you get when you are attracted to principle 1 and try to operationalize the principle. You imagine being justified in believing a claim. And you want to know what is true at the metalevel. So, instead of asking whether the information involved in the total perspective of the individual in question involves information sufficient to justify the claim that Jp, one resorts to operationalizing: imagine reflecting in such a way that total epistemic condition is preserved, and see what results one can defend. But such a move to reflection is precisely the distinction between being justified and being able to justify a given claim, and is a typical operational move to replace a concept with its operational consequences. My thought here is that evaluating access internalism should be one research project, and evaluating principle 1 a different one. I endorse the idea that undermining principle 1 will also undermine access internalism, but not vice-versa.


Access Internalism and JJ principles — 19 Comments

  1. Hey Jon,

    It’s late, so I might be missing something obvious but I thought I’d ask anyway. Can you say more about the cases that you think cause trouble for JJp entails Jp? You mention something about internal defeaters, but what sort of case did you have in mind? The ones that come most naturally to my mind are defeaters that defeat the justification for the first-order belief but thereby defeat the justification for the second-order belief. Those won’t do, we need defeaters that threaten the justificatory status of the first-order belief but leave the justificatory status of the second-order justification attributing belief untouched.

    At any rate, I can see lots of folks being attracted to 2. Suppose we say that p is justified for S iff it is reasonable for S to believe p. I can’t see how it could be reasonable to believe that it is reasonable to believe p and yet not be reasonable to believe p if, say, the subject is sufficiently appreciative of the fact that it is reasonable to believe that p is reasonable to believe.

    I think the same would hold true for blamelessness. If I blamelessly believe that I’m blameless in believing p, I can’t see how I could then be blamed for believing p.

    Or, consider your favorite view, the K-account of justified belief. (Where’s the tongue in cheek key?) You can’t know that you know p unless you know p.

    Or, consider a slightly better view than the K-account of justified belief. On the truths-reasonably-taken-to-be-true account, S’s belief that p is justified iff p and S reasonably believes it. If I satisfy the conditions for justifiably believing that I justifiably believe, that I’ve satisfied the antecedent of 2 entails that the consequent of 2 is satisfied as well.

    (Fwiw, I’m skeptical of 1. and know that there’s got to be a problem with 2., but it looks like the status of 2. might depend upon which account of justification you have in mind.)

  2. Hi Clayton, I wrote after a martini! Anyway, it was supposed to say “external defeaters”. I changed this before seeing your comment, but you are exactly right about that.

    So why reject 2? Well, the idea is what underlies my dissatisfaction with access internalism. Suppose you reflect the best you can and you are competent. But you get it wrong (after all, whatever factors plus principles make for justification, overlooking something is surely possible). You weren’t justified, but you’ve now come to the conclusion that you were and are justified.

    Here, the access internalist can say that the reflection itself now puts you in a different first-order state. Before reflection you weren’t justified, but now after reflection you are. I’m not convinced that is defensible, but I don’t have an argument against it either.

  3. Thanks, Jon, that clears that up. I think there’s something right about the rationale you sketch for rejecting 2, but, like you, I think I’m worried about the sort of response an access internalist might give. It seems weird (to me) to think that you could get a well-founded first-order belief from the justifiers that justify the second-order belief if there’s not already sufficiently good first-order justification as to make the justification provided by the second-order belief otiose.

  4. It seems to me that various people (and not only access internalists!) accept the following instance of (2):

    (2*) J(~Jp) entails ~Jp

    But, it also seems to me that if there are cases of genuine epistemic disagreement about the nature of justification itself, then (2*) can sometimes fail. What do people think about this?

  5. Sorry — my (2*) is not an instance of (2). It’s an instance of “factivity” of J (or J-collapse). So, it’s somewhat orthogonal to this post. But, I’m still curious what people think about it.

  6. Branden, been thinking about you, just cited your review of Bovens and Hartmann in a piece on coherentism, will send.

    But: on 2*:
    (ridiculous grammar included as an exercise for A-R grammarians…)

    I actually like it, but because I think of defeaters as most amendable to outrageous subjectivism: if you think you have a defeater, you do!

    I think lots are tempted by this position, but won’t endorse it in print. I suppose the reason is that thinking doesn’t make it so, but that’s an A rather than an E proposition (on the aristotelian square of opposition).

  7. Hi Jon,

    Can one reconcile acceptance of (2*) with your claim that the Church-Fitch paradox concerns only operators stronger than truth? (I’m assuming, of course, that J distributes as it needs to).

  8. Hi Aidan, not seeing the worry, can you explain? Remember I granted that iterated operators might raise the issue as well, as long as the iteration produced the correct half of idempotence. But I don’t see that here. If J~Jp entails ~Jp, and p is a conjunction of p&~Jp, then we get ~J(p&~Jp). Assume distributivity. Then you get ~Jp and ~J~Jp, right? But the latter doesn’t imply anything that contradicts ~Jp.

    So I think I’m missing something…

  9. It’s been a while since I thought about this stuff, and perhaps I’m just confused. But the thought was this. Suppose we’ve got p -> Jp around. Suppose also that J distributes over conjunction and obeys the reflection principle 2*. Doesn’t that give us a contradiction, by the usual reasoning (modulo replacing the appeal to factivity with an appeal to reflection)?

  10. Aidan, I goofed. So we assume J(p&~Jp). Then distribute, giving us Jp and J~Jp. By 2*, we get Jp and ~Jp. So we have to discharge the assumption, giving us ~J(p & ~Jp), a theorem, so we can box, and then take the dual, giving us that it isn’t possible to get J(p & ~Jp).

    So nice mimicking of the knowability paradox. But of course subject to the same approach I recommend, so long as we are starting from the obvious truth that there are some truths that aren’t justified. We get a contradiction with ” p implies possibly Jp” using exactly the same substitution rules that I don’t like in knowability.

  11. Thanks, guys!

    Jon — thanks for sending me your piece on coherentism — I look forward to that!

    I agree that many people seem to (at least, privately!) like (2*). But, it seems to me that (2*) must be false. I’ve been working on possible counterexamples. Let me think about this some more, and then maybe I’ll post a possible counterexample for discussion?

    Thanks, also, to you and Aidan for the connection to the Fitch paradox. Very neat!

  12. It occurs to me that Jon’s principle:

    (1) Jp entails JJp

    entails Branden’s principle:

    (2*) J~Jp entails ~Jp

    given the plausible assumption:

    (3) Jp entails ~J~p.

    A substitution instance of (3) is

    (4) JJp entails ~J~Jp

    And now from (1) and (4) we get

    (5) Jp entails ~J~Jp

    which is the contrapositive of Branden’s (2*). So, Branden’s (2*) is definitely not orthogonal to the issues raised in this post: if there are counterexamples, as Branden suspects, then we’ve got to give up (3) or, more likely, (1). As a fan of (1), I’ll be interested to think about Branden’s cases.

    By the way, a similar argument shows that from (1*) ~Jp entails J~Jp, we get Jon’s (2), i.e. JJp entails Jp. Jon, since you reject (2) does that mean you’re also willing to reject (1*)? Personally, I’m inclined to think that (1) and (1*) stand or fall together, but I’d be interested to hear if you think differently.

  13. Declan, yes, I don’t like (1*) either. I’d like us to be sensitive to the presence of first-order justification (that if it is there, there are also adequate grounds for the conclusion that it is there), but I don’t think that commits us to the claim that when first-order justification isn’t there, there are grounds for the conclusion that it isn’t there. It would be nice if that latter conclusion were true, and I’m interested in Branden’s counterexamples now more than ever.

    I like (3), but it is worth noting that it rules out epistemic dilemmas of a certain sort. I’m not inclined to use that as a reason to reject (3), but some might. I think Mike Bergmann is committed to rejecting (3) on this basis.

  14. Jon, here’s why I’m inclined to think (1) and (1*) stand or fall together. The first step is that if one lacks justification to believe that p, then one has justification either to disbelieve it or to suspend belief. The second step is that if justification for belief is self-intimating in the sense articulated by (1), then the same applies to justification for disbelief and suspension of belief. So, if one has justification to disbelieve or suspend belief in p, one has justification to believe that one does. The third step is that if one has justification to believe that one has justification to disbelieve or suspend belief in p, then one has justification to believe that one lacks justification to believe p. From these three steps, it follows that if one lacks justification to believe that p, then one has justification to believe that one lacks justification to believe p, i.e. ~Jp entails J~Jp. Thoughts?

  15. Yes, I balked at the second sentence. I think epistemologists ignore the distinction between taking no attitude whatsoever toward a proposition and withholding on that proposition. I think to take any attitude toward a proposition, one may need justification for doing so. But one may also take no attitude whatsoever, and may do so when there is no justification for taking any attitude. I don’t see why it would also be necessary that there be justification for taking no attitude.

  16. Jon, I’m happy with the distinction between the attitude of withholding and the absence of any attitude. But assuming I lack justification both for p and for ~p, is there a further distinction to be drawn between cases in which I have justification to withhold and those in which I lack justification to adopt any attitude? I would have thought I’d always have justification to withhold in this sort of case…

  17. Declan, I think not. To justify withholding, you need something to go on, and in many cases, we have absolutely nothing to go on. Absence of evidence shouldn’t be thought of as evidence for withholding. One way to model this in terms of degrees of belief is to consider the difference between a case where you have nothing to narrow the range of appropriate degrees of belief, leaving the range anywhere between 0 and 1 and cases where the range is narrowed to whatever the range is for withholding. I am wary of this way of modelling, since it looks like a confidence interval between 0 and 1 when one takes no attitude, and maybe there is an attitude which is a confidence level between 0 and 1. But narrower confidence ranges should be distinguished from taking no attitude at all. But the basic idea is that there is a confidence interval or intervals which count as withholding, and they contrain the interval in an important way, and that takes something epistemic to go on. Taking no attitude doesn’t take something to go on.

  18. Declan — nice point! Maybe this is why I thought (2*) was relevant. Anyhow, I’ll think more about the sorts of examples I have in mind (which have to do with disagreements about the nature of justification itself), and I’ll post something here soon (probably late next week, as I’m a little preoccupied until then).

Leave a Reply

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