Neta’s “An Internalist Refutation of Fallibilism” — 18 Comments

  1. Ram, rather than put up a whole list of questions, I’ll do some one at a time as they occur to me. The first is about your counterexample to Wedgwood’s claim about basic rules (and maybe Ralph will join in too if he’s not too lost in tutorials…). Ralph’s principle is:

    If one fact is partially constituted by a second, and a certain effect would still have been produced even if the second fact had obtained while the first fact had not, then if either fact explains the effect, it is the second fact rather than the first.

    Your counterexample involves Sam and Joe each believing that they have made inductive inferences, Sam by remembering them and Joe, not remembering, but having been told by a reliable source. Both believe as well that if they’ve made no inductive inferences, then their memories are bad. (It’s a material conditional, so Joe can infer it from the falsity of the antecedent.) You then say,

    So both Sam and Joe make the same transition in belief, but the state that explains Joe�s transition (his belief that he has made inductive inferences) is a constituent of the state that explains Sam�s transition (his forming his belief, on the basis of memory, that he has made inductive inferences).

    As I read the case, I think that the only facts that meet Wedgwood’s description involve something like Kaplanian character. So the “second fact” in Wedgwood’s principle is the Kaplanian character of Joe’s belief, and the “first fact” is, perhaps, Sam’s belief, which includes the character in question. Construed this way, the counterexample would need to claim that it is false that it is the shared character, rather than Sam’s belief, that explains the transition in question (since it is obvious that something in the neighborhood is explanatory).

    Is this the way the counterexample is meant to be understood? If so, I’m having trouble with the clause “even if the second fact had obtained while the first fact had not”. If the facts in question are Kaplanian characters, then this clause can’t be satisfied, and even if Ralph’s principle doesn’t require this possibility, it would be an obvious amendment to it (to rule out the counterfactual being trivially satisfied when the antecedent is impossible). If the fact is not a Kaplanian character, but rather the character of Joe’s belief, then it’s not a constituent of Sam’s belief.

    Am I misreading here?

  2. Ram, I ought to own up to some assumptions in the worry in the second-to-last paragraph. First, the second fact, the Kaplanian character, depends only on the indexical chararacter of the beliefs in question (not the rest of the relevant sentences). Second, the identity conditions on beliefs requires an item that explains the causal role beliefs play, and the indexical character of the belief is thus essential to its causal role. Finally, something about indexical character: the relevant function puts the same content of belief in Sam’s belief box in any world in which the indexical character is a constituent of a belief state of Sam’s. So there can’t be this particular character in Sam’s head without this particular content there as well. (This last one is meant to clarify the impossibility I’m worried about–it is possible for this character to be in some other head and Sam’s belief state not to obtain, but it is not possible for this character to be in Sam’s belief box and this belief state not to obtain, as long as the distinctive function that is the character in question depends only on the indexical nature of the sentences in question).

    I hope this helps make clear how I’m thinking about the case, and may make it easier to say how you’re thinking of the case that I’m missing…

  3. Hi Jon,

    Thank you for posting my paper, reading my paper, and pressing me on this important issue concerning Ralph’s principle.

    Now, given the way that you think of the identity conditions on belief, my counterexample to Ralph isn’t going to work. And that’s too bad for me, because I was hoping that I could avoid getting into the complicated issue of what the identity conditions for beliefs are.

    But I still think that there’s something intuitively plausible about the counterexample — maybe plausible enough to carry some weight against a view that would individuate beliefs as finely as the view you suggest….?

    Let me try to bring this out a bit more clearly by saying what I took to be the cause-effect pairs in the case of Sam, and in the case of Joe:

    Sam —
    Cause (has two parts):
    (a) believing that he’s made inductive inferences, and
    (b) believing that his belief that he’s made inductive inferences is based on his memory

    believing that IF he has not made inductive inferences, THEN his memory is bad

    (Also, let’s suppose that, in the case of Sam, the two-part cause specified above directly gives rise to the effect specified above, without the intervention of any further beliefs.)

    Joe —
    believing that he’s made inductive inferences

    believing that IF he has not made inductive inferences, THEN his memory is bad

    (Let’s suppose that Joe is extraordinarily good at simply inferring whatever follows from his beliefs even though he doesn’t have any beliefs about logic. So, in the case of Joe, the cause specified above directly gives rise to the effect specified above, without the intervention of any other beliefs.)

    So, the same effect is produced in the cases of Sam and Joe. (Or at least, the effect is of the same kind, adjusting for differences in the referent of “I”.) And the cause of that effect in the case of Joe is a constituent of the cause of the effect in the case of Sam. (Again, adjusting for differences in the referent of “I”.)

    Given your assumptions about belief-identity, you will not accept this conclusion. But, if it’s possible to put aside general views about belief-identity, isn’t there something plausible about my claim that the cause in the case of Joe is a constituent of the cause in the case of Sam? (I know that this isn’t really an argument. But I’m still hoping that we can evaluate Ralph’s principle without having a theory of identity conditions on belief.)

  4. No, I’m not lost in tutorials at all — I’m on sabbatical (until the end of September), and I’ve taken myself off into quasi-monastic seclusion in Princeton, New Jersey, to try to get work done (mostly on metaethics rather than epistemology, I’m afraid).

    Anyway, thank you so much for all your attention to Wedgwood (2002)! No doubt you will not be surprised to hear, however, that I deny that Ram has produced a counterexample to my principle.

    The effect that (according to Ram) counts as a counterexample to my principle is this: Sam’s coming to believe that if has has not made inductive inferences then his memory is bad. However, Ram’s example as currently constructed does not involve the assuption that if Sam had merely believed that he has made inductive inferences (and not also believed that he holds this belief on the basis of memory), the effect in question would still have come about (i.e. Sam would still have believed that if he has not made inductive inferences, his memory is bad). But unless Ram’s example includes this assumption, it will not even begin to be a counterexample to my principle.

    But, perhaps we could just add this assumption to Ram’s example? Sam would have reasoned just as Joe in fact does, but his reasoning in this way is pre-empted by his being caused to reason to the very same conclusion in this other way.

    In fact, when I try to make the reasoning that Sam actually goes through intelligible to myself, I always find myself imagining that in this reasoning, the fact that Sam believes that he has made inductive inferences actually plays no causal role at all. He believes that his memory tells him that he has made inductive inferences, and this belief alone is enough to cause him to believe that if he has not made inductive inferences, then his memory is bad. After all, he surely could reason in this way even if he temporarily suspended judgment completely about whether or not he really has made inductive inferences. So in this reasoning — at least if it is a process of rational reasoning, and does not involve the reasoner’s being swayed by evidentially irrelevant considerations — the cause of the belief in question does not in fact have any constituent in common with what would be the cause if Sam went through the other sort of reasoning instead.

    Well, but what if Sam is being irrational in this way, and this irrational reasoning preempts the rational reasoning (which is just like Joe’s reasoning) that he would have gone through? Perhaps this case, unlike Ram’s case, really would be a counterexample to my principle. So I may have to refine my principle.

    In fact, I am now much less enthusiastic than I used to be about appealing to counterfactuals, and more prone to try to appeal to dispositions (which I regard as closely akin to ceteris paribus laws or regularities) instead. So perhaps what really tells against appealing to a factive mental state, like knowledge, in giving the proximate folk-psychological explanation of a belief revision, is that the disposition that links the factive state to the belief revision in question is just a special case of the more general disposition that links the corresponding non-factive state to that belief revision. That is, the cases in which this first (factive-state-involving) disposition is manifested are properly included in the cases in which the second (non-factive-state-involving) disposition is manifested. So we achieve a more general explanation by appealing to the non-factive mental state instead of the factive state, without losing any explanatory value.

  5. Hooray for the move to dispositions from counterfactuals, Ralph! One way to curb a penchant for finding Shope fallacies is to try to read appeals to counterfactuals in terms of dispositions realistically construed (i.e., not analyzable in terms of counterfactuals).

    And jealous of your sabbatical…

  6. Hi Ralph,

    Thank you very much for your response. I’m still puzzled about something.

    To spell out what puzzles me, let me start by labeling some propositions.
    p = if I’ve made inductive inferences, then my memory is bad
    q = my memory tells me that I’ve made inductive inferences
    r = I’ve made inductive inferences

    As I’m envisioning the case of Sam, the cause of his belief that p has two parts — his belief that q and his belief that r — and neither of these two parts alone is (in the case I’m envisioning) causally sufficient for the effect. In other words, I’m supposing that Sam just wouldn’t infer from his belief that q to his belief that p unless he also believed that r.

    Now of course, there are other possible cases. For instance, as you correctly point out, there’s a possible case in which Sam’s belief that p is produced just by his belief that q, without the cooperation of his belief that r. (In fact, that would be a much more ordinary case.) But I was trying to describe a case in which the two causes (his belief that q and his belief that r) are causally sufficient only in tandem. If such a case is indeed possible, then it is a counterexample to your principle, no? And if it is a counterexample to your principle, then you need to deny that the kind of case that I’m trying to describe is possible. Is that right?

    So do you want to deny that there’s a possible case of the sort that I’m trying to describe — one in which Sam’s belief that p is caused jointly by his belief that p and his belief that q, but neither is alone causally sufficient?

    Of course, I haven’t said anything here about your general remarks about dispositions connecting factive states with belief-revisions. I’d be interested to talk about that also, but first I’d like to get clear on the more specific issue concerning Sam…

  7. Dear Ram,

    Thank you so much for your reply. You ask:

    But I was trying to describe a case in which the two causes (his belief that q and his belief that r) are causally sufficient only in tandem. If such a case is indeed possible, then it is a counterexample to your principle, no?

    Well, no — at least strictly speaking, it’s not yet a counterexample. To get a counterexample, we have to add the assumption that the effect in question would still have come about if the belief that r had been present but the belief that q had not. The effect in question is Sam’s coming to believe p. So the assumption that we need to add is that Sam would have reasoned in precisely the same way as Joe if he (Sam) had believed just r and not q as well. As I concede, however, it looks as though we can consistently add this assumption to your example.

    I didn’t deny that the causal process that leads Sam to come to believe p was possible, only that it didn’t count as a rational process of reasoning. But since my original principle was not restricted to cases to rational reasoning, this point is probably irrelevant. This is why my final suggestion was that that I may need to revise my principle so that it is formulated in terms of dispositions instead of counterfactuals.

  8. Ralph–
    You say:

    That is, the cases in which this first (factive-state-involving) disposition is manifested are properly included in the cases in which the second (non-factive-state-involving) disposition is manifested. So we achieve a more general explanation by appealing to the non-factive mental state instead of the factive state, without losing any explanatory value.

    But is it always going to be the case that the more general explanation has more explanatory power?

    That question may commit Jon’s philosopher’s fallacy, but my worry is this: We are considering cases in which the disposition is actualized–not only is Sam disposed to form beliefs under certain circumstances, those circumstances obtain and he actually forms those beliefs. Ram has set up the example (I think) so that the disposition to move from (q and r) to p causally preempts the disposition to move from r to p. Then, when we’re trying to explain why Sam believes that p, doesn’t the two-part belief (q and r) have more explanatory power than the one-part belief r, since the belief (q and r) is the actual cause of the belief p?

    To make things clear, let me make these stipulations (and see if everyone finds them acceptable):

    Sam is disposed to form a belief that p two seconds after he forms a belief that r.
    Sam is disposed to form a belief that p one second after he forms a belief that (q and r).
    At 12:00:00 Sam forms the belief that (q and r).
    At 12:00:01 Sam forms the belief that p.
    So we can tell that the second disposition is the real cause.

    Perhaps your idea that dispositions are regularities prevents us from saying that one preempts another?

  9. Ram, I must admit to being a bit nervous about your premise of Basicness Uniformity. You claim that it is “a corollary of the general claim that basic knowledge is not, in general, a greater or more demanding accomplishment than non-basic knowledge”. But there are many ways in which a demand might be greater or more demanding. In particular, the paradigm cases of basic knowledge are ones in which the agent is not called upon to do very much, by way of offering extra-good reasons for their belief, etc. So in that sense, basic knowledge is less demanding than non-basic knowledge. But not just in spite of that fact, but because of it, there might be other conditions on basic knowledge that are stricter than the conditions on non-basic knowledge. Basic knowledge cases are exactly those in which you’re relying on a source of evidence without any particular evidence that it’s a reliable source of evidence to be relying on here — it makes some sense to me that we might require that, in these cases, the source turn out in the case to be maximally reliable.

    Note that imposing an infallibility condition on basic knowledge is a stricter condition than we might have on non-basic knowledge, but not obviously more strenuous for the agent. In particular, if you’re generally right about the generally easy availability of some forms of infallible justification, like seeing, then an infallibility condition won’t be particularly demanding to meet. It might turn out that it’s a condition that is only imposed rarely, but also only imposed in circumstances where it’s not too hard to meet.

  10. Dear Ralph and Matt,

    Very helpful points! I’ve modified my example considerably, and now I think I may have come up with a counterexample to Ralph’s principle, even given the modification suggested above. If you have a chance, Ralph, I’d be very interested to hear what you’d say about this case…

    Sam has a compelling apparent memory of having made inferences. This compelling apparent memory is a complex psychological state, one constituent of which is Sam’s confident belief that he has made inferences. (Without that belief, the apparent memory wouldn’t be compelling.) Now, on the basis of his compelling apparent memory of having made inferences, Sam rationally comes to believe that if he has not made inferences, then his memory is deceptive (i.e., not just defective, but productive of false beliefs). Sam’s belief in the conditional is caused by the complex state of his compelling apparent memory of having made inferences: a complex state one constituent of which is his belief that he’s made inferences. This complex cause directly and rationally gives rise to the effect of Sam’s believing that if he’s not made inferences, then his memory is deceptive.

    But suppose that Sam had simply believed that he had made inferences, and had not believed anything about his memory. In that case, this belief would also all by itself have rationally caused him to believe (by inferring it from the falsity of the antecedent) that if he has not made inferences, then his memory is deceptive.

    So either way, Sam is rationally caused to believe that if he’s made inferences, then his memory is deceptive. And the belief that rationally causes this effect in one case is a constituent of the complex state that rationally causes the effect in the other case.

    Is this a counterexample to the principle, even as amended?

  11. Hi Jonathan,

    Thanks for reading the paper! Your post and my last post must just have crossed each other in cyber-space. I hope that this post doesn’t suffer from the mysterious typographical strangeness of my last one.

    You’re right that I don’t do much to motivate basicness uniformity. Your rationale for worrying about it sounds right to me as well. And worse yet, if basicness uniformity is wrong, then my internalist argument for infallibilism about knowledge is unsound.

    But notice that, even if I concede all this, I’d still have an internalist refutation of fallibilism about basic knowledge. And that would be fine by me, especially because it would motivate the point that I really care about the most, which is that it’s possible to have reflectively access to infallible grounds.

    By the way, on the topic of basic empirical knowledge, Ted Poston has very astutely brought it to my attention that it seems unclear how there could be basic empirical knowledge if closure is true. If I have empirical knowledge that there’s a sheep in front of me, and I have reflective knowledge that it appears to me as if there’s a sheep in front of me, and if conjunction closure is true, then don’t I have to be able to know, by deduction, that my experience is veridical? How is it so much possible then for there to be basic empirical knowledge?

    In reply to this question, I should say that I think that knowledge of the veridicality of your experiences cannot be gained by this kind of conjunction deduction. But Closure doesn’t tell us anything about how knowledge is acquired, or gained. It tells us only what knowledge needs to be present under what circumstances. So if you know that there’s a sheep in front of you, and you know that it appears to you as if there’s a sheep in front of you, and you make the right kind of competent deductions, then you also know that your experience is veridical. But it doesn’t follow from this that you acquired knowledge about the veridicality of your experience by making those deductions. Nothing follows about how you acquired the knowledge. I need to revise the paper to make this point much clearer.

  12. Matt: I certainly don’t want to deny that the manifestation of one disposition can pre-empt the manifestation of another disposition!

    Your stipulations about the case are absolutely fine. I was in fact assuming that in the amended version of Ram’s case the disposition that takes Sam from (believing q and believing r) to believing p pre-empts the disposition that would have taken Sam from believing r to believing p.

    Still, I deny that this is a counterexample to the amended version of my principle because these two dispositions are not related to each other as a more general law is a related to a more specific version of that general law. (I.e., they are two different kinds of reasoning dispositions: they are not related to each other as genus and species.) Of course, I haven’t said what the dispositions in question are. But intuitively, one is a disposition for one sort of inference (inferences of the form “Not-p: Therefore, If p then q”), while the other is a disposition for an entirely different sort of inference.

  13. Ram: the issue of the rationality of Sam’s reasoning is probably a red herring (since neither my original principle, nor the amended version that appeals to dispositions instead of counterfactuals is restricted to rational belief revisions); I’m sorry I brought it up.

    For the record, though, I might as well mention that I’m not convinced by your attempt to make Sam’s reasoning rational by appealing to this ‘complex psychological state’. I’m inclined to think that the relevant causes and effects are not token states, but rather facts (like the fact that Sam came at t to be in a mental state of such-and-such a type).

    So, even if Sam’s token mental state is “complex” in the way that you describe, it still entails that Sam simultaneously has the mental state type “seeming to remember making inductive inferences” — where this type (unlike the complex token state) is not belief-entailing. I claim that the disposition that Sam manifests, if he is rational, is a disposition that responds to his being in the non-belief-entailing memory state type. This is because the fact that he is actually convinced by his memory that p is epistemically quite irrelevant to supporting the conclusion that if not-p then his memory is bad.

    As I say, however, I think this point is probably a complete red herring. I would respond to your counterexample in the way that I outline in responding to Matt above.

    If the two dispositions are related as species and genus, on the other hand, then I deny that your description of the case really is incoherent. It cannot be that both believing q and believing r are causally necessary to cause Sam to believe p, and at the same time believing r alone is causally sufficient to cause him to believe p. So the only reason for saying that the cause is Sam’s both believing q and believing r is that this cause leads to his believing p via a different sort of process from the one that would have led to his believing p from his believing r alone. What makes it a different sort of process is that the disposition that is manifested is of a profoundly different kind, not just a species of the very same disposition that would have been manifested if Sam had reasoned in the way that Joe did.

  14. Ralph–I’ll have to think about it more. I won’t be at all unhappy if your argument does go through!

    Ram–I am worried about your argument for No Divination against the advocates of contingent a priori knowledge.

    (1) You say (p. 14) of the CAPK theorists that “Their arguments all rest on some contingent premise that can be known only empirically, e.g. that we have empirical knowledge.” But mightn’t one hold that “We have empirical knowledge” in particular can be known a priori? It might be the foundation of our synthetic a priori knowlege (don’t take my invocation of Kant too seriously). It’s not obvious to me that “we have empirical knowledge” can only be known empirically. Perhaps reflection will reveal that any creature that is capable of thought must have some empirical knowledge, or something like that.

    (2) Here’s an example of one of the things that, it is claimed, might be known contingent a priori: “If [in fact] it has snowed in Ithaca every winter for the past 200 years, and [in fact] Ithaca is not a marginally wintry place, and [in fact] there is no reason to believe that the earth will get considerably warmer in the next year, and [in fact] we have found that weather patterns obey regularities, then [in fact] it will snow in Ithaca sometime during the next winter [in particular].” I’ve tossed in all the bracketed “in facts” to make clear that it’s not a general law or induction principle that’s supposed to be known a priori–it’s an instance of this generalization. Why must knowledge of this rest on a contingent principle that can only be known a priori? I think CAPK theorists might argue that we could arrive at this sort of thought by reflection alone, and that when we arrive at this thought it does constitute knowledge. [If it turned out to be false it wouldn’t constitute knowledge. But it’s not going to.] The bases for this reflection might be rules for induction and the like, but it’s not obvious that these rules could be known only empirically. It’s not obvious that the rules are even contingent. Heck, I suspect that the principles can’t be supported empirically.

    (3) I suspect that you have in mind something like this: The reason that CAPK theorists want there to be contingent a priori knowledge is something like this: They reason that if there isn’t CAPK, there can’t be empirical knowledge. And there’s empirical knowledge. So there’d better be CAPK.

    Well, that seems fair–that is how the papers go if I remember correctly. But that doesn’t show that the CAPK itself can only be supported by the premise that there’s empirical knowledge. The CAPK may be supported by other factors, as in (2). I think, the argument from “there is empirical knowledge” is meant to bolster the notion that these cases should count as knowledge–it’s not part of the reasoning that gives us this knowledge.

    (Thanks for the very neat paper!)

  15. Hey Matt: Thank you for reading my paper, and for the interesting comments.

    Can you know a priori that you have empirical knowledge? Suppose you were put into a complete sensory deprivation tank (no sensory perception or proprioception) and then given an amnesiac. Would you then have empirical knowledge? It seems to me that you wouldn’t — all your knowledge would then be reflective knowledge of your own thoughts, images, desires, etc. Now, how do you know that you’re not in that situation right now? How do you know that the various psychological episodes of which you’re currently aware are not themselves merely images, thoughts, etc.? Don’t you know that empirically? How do you know — if not empirically — that a particular psychological episode is a perceptual experience, and not merely an image conjured up by imagination?

  16. I think the line I’m entertaining in #1 is that a person in such a situation wouldn’t be able to know anything whatsoever, reflective or not. So, from the fact that I have reflective knowledge I can conclude by reflection that I have empirical knowledge.

    I guess I don’t feel the force of your last question because it’s not obvious to me that empirical knowledge helps much here. You give an explanation of how we can know this empirically, but someone who thinks that CAPK is easier to accept than your picture may be fine with saying that the general premises from which all things spring–“I have empirical knowledge,” “I am not a BiV,” like that–are known non-empirically. Knowledge that this episode is a perceptual experience will still be empirical, too; it’s “if I seem to perceive p [blah blah ceterus paribus], then p” that’s known non-empirically. The antecedent is supplied by experience, so p itself is known empirically, and “This experience as of p is veridical” is also known empirically.

  17. Hey Matt: An internalist — on my definition — will grant that S knows a priori that p only if S has reflectively accessible a priori grounds to believe that p. I guess what I want to hear from the defender of CAPK is: what reflectively accessible a priori grounds do I have to believe that “if I seem to perceive p, ceteris paribus, then p”?

    I guess that Hawthorne and Weatherson aren’t concerned to answer this question because (I think) they aren’t internalists. But I’m just addressing internalists here. And so I’m wondering how they might wish to answer the question above on behalf of the CAPK theorist. In particular, I’m puzzled by BonJour’s adherence to CAPK, because I don’t see that he does answer the question I’m posing to the internalist. Have I missed something in BonJour?

  18. I can’t speak for BonJour on this, but I’ve just looked back at the end of In Defense of Pure Reason. He says that for a conclusion to follow rationally (certainly or probably) from premises is for “[t]he connection between premises and conclusion [to] be, one might say, intellectually visible” (p. 203). He then goes on to justify induction a priori by arguing, more or less, that it’s obvious that if a regularity of a certain kind has been observed, it’s very unlikely that that regularity is mere coincidence (p. 208). And maybe he’d say something similar about perception–it’s just obvious that, if you seem to perceive something, the best explanation is that it’s out there (perhaps this should be applied to the whole world: it’s obvious that, if you’ve had lots of perceptual experiences of the world, and they hang together in the way that they do, the best explanation is that the world is out there–then you can get to instances of “If I seem to perceive p, then p” by instantiating.) But as I said, I can’t speak for him.

    Another person to think about here is Burge. Burge may be a semantic externalist, but mayn’t he intend his acceptance principle to be reflectively accessible apriori? He justifies it with arguments about the nature of reason, etc., and those seem meant to be reflectively accessible and apriori.

    I’m not sure I’ve succeeded in actually pointing out specific reflectively accessible grounds for CAPK–but I do think that there are people who think that there are reflectively accessible grounds.

Leave a Reply

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