Neta on Naturalized Epistemology

Ram has a new paper called How to Naturalize Epistemology. I’ll put it on the “Works in Progress” page, and discussion is welcome here or separately to Ram. It’s good to see someone got something done this summer!

UPDATE: As of 8/9 at about 8:10 pm, there is a new version of the paper, using the same link above and the link in the Work in Progress page.

ANOTHER UPDATE: As of 9/6 at 9:40 am, there is another new version of the paper in place, same links as before


Comments

Neta on Naturalized Epistemology — 78 Comments

  1. Looks like a neat paper! But, early on, Ram assumes “E confirms H more highly than H’ if and only if Pr(H/E) > Pr(H’/E)”. I’ll have to read the whole paper carefully before seeing if this makes sense in context (perhaps by “confirms” Ram doesn’t mean “is better evidence for”?). But, I have argued elsewhere that this is not a very good probabilistic approach to relational confirmation. See:

    http://fitelson.org/synthese.pdf

    This issue was also discussed on Certain Doubts (by Jon), at:

    http://bengal-ng.missouri.edu/~kvanvigj/certain_doubts/?p=141

  2. It seems reasonably clear that it is not in general true that E confirms H more than it does H’ iff. Pr(H/E) > Pr(H’/E). In fact, it is not possible to tell (or else, I certainly can’t tell) how much E confirms H from that inequality. The obvious counterexample in (1)-(3),
    1. Pr(H/E) = Pr(H), and

    2. Pr(H’/E)> Pr(H’), and

    3. Pr(H/E) > Pr(H’/E)

    I can’t imagine claiming in spite of (1)-(3) that E confirms H more than it does H’. E does not confirm H at all! Nor does it provide any evidence for H (that we don’t already have). And cases where (1)-(3) are true are easy to construct. It would be great to know what argument there is that in general E provides more evidence for H than H’ given (just) that inequality. Maybe Branden knows what argument(s) have been offered. On the other hand, I’ve still to finish Ram’s paper (maybe the argument is there).

  3. I’m finding myself unable to open the paper at the Work in Progress page–I’m getting some sort of WordPad error. Is this unique to my computer? More important to me: Does anyone know of a way around this difficulty?

    Keith Brian Johnson

  4. Ram, as you can probably tell by my Kornblith quote, I read your paper and liked it very much! I don’t have any objections to it, but a couple of thoughts might be useful. First, I always worry when the concepts of a goal and a function and other teleological notions get used, since I’m pretty sure these aren’t the same concepts. Notice, for example, in section 3 that you adopt a straightforwardly Plantingian account of epistemology, but then when you characterize your own view, you don’t talk about the function of a cognitive system, but rather about the way the organism uses a cognitive system to achieve its goals. The latter might be the usual resort to intentional language to describe what can be described in purely biological language, but that strikes me as needing an argument. Besides, even if we had one, then the focus is not on the cognitive system itself, but on the larger system that incorporates cognitive and practical modules to generate some degree of inquisitiveness on the part of the organism. It’s not clear to me how this comports with your objection to B&T that practical matters shouldn’t infect the epistemology.

    I agree with Branden’s concern about the notion of confirmation being used, and I think it is relatively easy to adjust the discussion to take Branden’s beautiful paper into account. But the use of confirmation here raises a further concern, one in the area of my post today about Kornblith and the importance of knowledge. If knowledge really was all that important, then what we should expect is an explanation of how we know that your own teleological naturalism is superior to B&T and K. But we don’t get this. We get instead a confirmational argument, which I applaud. If that’s what really counts, however, it is not clear why we are spending so much time trying to naturalize knowledge. (Can you detect some value-driven considerations here?!) The action would all be in trying to naturalize confirmation instead, wouldn’t it?

  5. Hey guys,

    Thanks for reading my paper so quickly, and for the comments!

    Brandon and Mike: I now believe that the word “confirm” should be banned from all future philosophy. It is used in too many different ways to be useful anymore. Let me restate the assumption of my paper without using the term “confirm”.

    Consider an epistemic agent S, whose total evidence at time t is TE. Now consider some hypotheses h1, h2, h3, etc. Question: at time t, how should S distribute her confidence over the various hypotheses? Here’s my assumed answer: S’s degree of confidence in hn should be proportional to Pr(Hn/TE). That’s the assumption that my paper employs. And from now on, thanks to you guys, I’m going to state that assumption without using the term “confirm”.

    (Brandon, I look forward to reading the paper to which you refer me!)

    Now, Jon: You’re absolutely right that goal and function are different, and I was not paying attention to the difference in my discussion. On my view, the function of our cognitive system is to achieve a particular goal (namely, knowledge). But the function is not the same as the goal, and I didn’t make that clear.

    Do I want to naturalize confirmation? Actually, I want to throw away the term “confirm”, so I can’t answer that question.

  6. Ram, do you think there is an argument for the claim that knowledge is the goal of our congitive system or systems? Why not, for example, truth; or understanding?

    I’m not sure what to make of the confirmation point anymore, but there’s still something perplexing about not arguing that we know, in terms of the account of knowledge you prefer, that TN is superior. Whatever distinct terminology you use, you’ll generate the need for more naturalizing. I’m not saying that’s debilitating; just noticing it.

  7. Hi Jon,

    I think there is an argument for the claim that knowledge is the goal of our cognitive system. I just don’t know what the argument is.

    Probably, the argument involves reflecting on what it is that would satisfy you, if you were inquiring into the issue whether p. Would you be satisfied with having a true belief as to whether p, without knowing that your belief is true?

  8. Ram, let’s distinguish between what satisfies you in inquiry, and what labels one would use to describe what satisfies you in inquiry. The latter has the problem that ‘knowledge’ has honorific content in addition to what it really is. And when we substitute the real nature, according to Kornblith’s theory, for the label ‘knowledge’, we won’t be able to get that it is knowledge (i.e., reliable true belief) that satisfies, at least not for those who realize that reliability adds nothing to the value of true belief.

    There is still the question of what sates the inquisitive attitude, and here I think it is less plausible to think that it is knowledge that does this. Satiation occurs when the person experiences closure with respect to the issue at hand, and though there may be a rough correlation between that experience and having acquired knowledge, that’s all there will be. And that correlation depends on a cooperative environment. It’s easy to see how to design actual experiments where people experience closure, but, unknown to them, the environment is not cooperative and hence they lack knowledge.

  9. Thanks for the clarification, Ram. No need to banish “confirmation” from philosophy! At least, let’s wait until I have tenure before we do that. What you are talking about is what Carnap called “confirmation as firmness” (as opposed to what he called “confirmation as increase in firmness,” which is what contemporary Bayesians are talking about — incluing me the paper I cited above). That is, you are talking about CREDENCE, not confirmation. Given that reading, what you say makes more sense. But, I would NOT characterize “Pr(H | E) > Pr(H’ | E)” as capturing the relation “E is better evidence for H than H'”. And, you don’t need to do that in your paper, as far as I can tell.

  10. Brandon: many thanks for the useful terminology from Carnap! I’ll appeal explicitly to it in the revision.

    Jon: on second thought, it was careless of me to say that what matters is what would satisfy you if you’re inquiring into the issue whether p. Any number of things might satisfy you, depending on your other (extra-cognitive) goals. The question really is: what is the function of our cognitive system? And how do we answer questions like that? How do we learn about the function of other biological systems (e.g., the circulatory system)? How do we learn about the function of artificial systems (e.g, political institutions)? We find out about these things somehow or other empirically. But I guess all we can do is compare various proposed hypotheses against the total evidence, whatever that may include. So I guess that’s not very helpful, but at least I’ve argued in my paper that my own view scores more points than Kornblith’s, or Bishop and Trout’s. And if some other theory scores more points than mine, then great: let me hear about the other theory.

  11. Ram, you say,

    “Here’s my assumed answer: S’s degree of confidence in hn should be proportional to Pr(Hn/TE)”.

    One’s degree of confidence in Hn does sound like one’s credence in Hn. But there are interesting situations in which this does not seem to be true. I am assuming that TE is one’s total evidence as of a certain time. It can for instance be true that TE is my total evidence for/against Hn at t and certain that, at t+1, TE is going to be modified. I’m on my death bed and TE has led me to conclude that Hn = God does not exist. At t+1 there will certainly be additional evidence E+ for/against Hn (whether or not I am in a position to learn that evidence at t+1). Practical considerations aside, how confident should I be on TE that I am right about Hn? After all, E+ is certainly forthcoming. More importantly, TE provides no evidence at all that E+ will favor Hn. None. So I wonder whether my degree of confidence in Hn should be proportional to TE (i.e. my total evidence at a given time) for/against Hn. Shouldn’t my degree of confidence in Hn depend in part on how sure I am that additional evidence for/against Hn is about to be made available?

  12. Hi Mike,

    I’m having a bit of trouble following. Maybe you can help me out a bit more on this point.

    Let G and H be two competing hypotheses.

    Let H’ = the hypothesis that a moment from now I will receive evidence that favors G over H.

    Let G’ = the hypothesis that a moment from now I will receive evidence that favors H over G.

    Let H” = the conjunction of H and H’

    And let G” = the conjunction of G and G’.

    Now, is there a situation in which Pr (H/TE) is higher than Pr (G/TE), and Pr (H”/TE) is not higher than Pr (G”/TE)? Is that the kind of situation that you have in mind when you say that it’s not clear that one should apportion one’s degree of confidence in H to Pr(H/one’s total evidence at the moment)? Do you think that there are situations in which Pr (H/TE) > Pr (G/TE), and yet one should be more confident of G than of H?

  13. “Now, is there a situation in which Pr (H/TE) is higher than Pr (G/TE), and Pr (H”/TE) is not higher than Pr (G”/TE)?”

    I’d make a slight modification, but it is probably what you meant anyway. I’d say that there are cases in which Pr(H/TE) is higher than Pr(G/TE) and Pr(H”/TE) is higher than than Pr(G”/TE). TE = I flip a coin twice and it comes up “heads” and then “tails”. Let H = it’s fair coin, G= it ain’t. Add to TE this = the chances are high that tomorrow there will be much more evidence for G. So Pr(H/TE) > Pr(G/TE) (i.e., the two coin flips would lead me to conclude that there is more evidence that the coin is fair than that it isn’t). But I also have reason to believe that my evidence tomorrow is going to favor G. So Pr(H”/TE) > Pr(G”/TE) (i.e. the evidence I have favors H over G and it also favors that I will tomorrow learn something favoring G)

  14. Wait a second. If we add to TE that the chances are high that tomorrow there will be much more evidence for G, then isn’t my new total evidence (call it TE’) such that:

    Pr(G/TE’) > Pr(H/TE’)?

    Or at least, it’s not the case that

    Pr(G/TE’)

  15. In ‘How Truth Governs Belief,’ Phil. Review, October 2003, I argue that accounts of belief in terms of the function of the system that produces them won’t explain the phenomenon I call transparency, which is that the deliberative question whether to believe that p gets answered by, and only by, answering the question whether p. In that paper, and ‘Doxastic Deliberation,’ co-authored with David Velleman (available on his website), I also suggest an alternative account of the normative relation between belief and truth that can explain transparency.

  16. “Wait a second. If we add to TE that the chances are high that tomorrow there will be much more evidence for G, then isn’t my new total evidence (call it TE’) such that:
    Pr(G/TE’) > Pr(H/TE’)?”

    Ram,
    I don’t think so. You don’t have the forthcoming evidence yet. All you have is a certain chance that tomorrow there will be such evidence. But, to avoid updating on uncertain evidence, let me exclude talk of chances. Put the case this way. Suppose the *total evidence* you have favors H and none of the evidence you have is relevant to G (e.g., you have evidence that pill A caused the problem and you have no evidence about pill B). Suppose tomorrow there will be lots of evidence relevant to G (pill B). It is true that Pr(A /TE) > Pr(B/TE), right? But I’m not sure my belief that pill A caused the problem should be proportional to TE. That would be a little rash, wouldn’t it?
    When you say “total evidence” it sounds like you’ve got all of the evidence for and against the relevant hypotheses. But this is not what you mean, I’m sure. You mean all of the evidence currently available. But it seems like my belief should not be proportionate to TE if I know (for instance) that TE includes no evidence relevant to one of the competing hypotheses and that such evidence is forthcoming.

  17. Hi Nishi,

    Yes, I’ve been intending to read your paper for some time now, and I hope to get to it soon. But I’m not sure that there’s going to be any incompatibility between what you say in that paper and what I say in my paper. I don’t claim to explain transparency, nor do I claim to offer an account of belief. (Although I think it would be good to do both.) I do, however, claim to offer an account of knowledge in terms of the function of the system that produces knowledge. Nothing in the introduction to your paper led me to believe that what you say is incompatible with that, but I look forward to reading and finding out.

    Mike, I think there’s one main point at issue between us. It’s this: You say that credence in H at a certain time should not be proportionate to the total evidence I have at that time, and that’s because I can know that more relevant evidence is forthcoming, and that the forthcoming evidence will alter the degree to which the hypothesis is supported. Whereas I think that the latter piece of knowledge is PART of your total evidence, and once it is taken into account as part of your total evidence, then you should still apportion your credence in the hypothesis at that time to your total evidence at that time. Now, I’m having trouble seeing what would support either of our views against the other, short of some general account of evidence.

  18. Hi Ram,

    Maybe you are right that there is no incompatibility between your claims about knowledge and my claims about belief, but, from a brief glimpse at the first three sections of your paper, I notice that in section three you say,

    ‘I propose that we think of epistemic norms as norms that govern one of the functional systems – more specifically, our cognitive system – that are implemented by our nervous system. We can find out about these epistemic norms in something like the same way that we would go about finding out about the proper function of any of our other functional systems.’

    I would have thought that epistemic norms most fundamentally are the norms we follow in determining whether to believe that p. If this question––whether to believe that p–– is governed by transparency, then epistemic norms will have to be norms for determining whether p. So it is only if the best explanation of transparency is in terms of the function of our cognitive system that we ought to accept your claim that epistemic norms are norms that govern our cognitive systems.

  19. Hi Nishi,

    Interesting! But there’s something I’m still not getting. I grant that epistemic norms are norms for determining whether p. And I also grant that epistemic norms are norms governing one of our functional systems (namely, our cognitive system). How does it follow from those two claims that I am committed to explaining transparency in terms of the function of our cognitive system? Sorry for being slow…

  20. Hi Ram! A couple of first-read questions/concerns:

    (i) Why have you embraced this version of naturalish epistemology, as opposed to the Craig-type that you & I have talked about in the past? Or do you think that they are complementary methodologies?

    (ii) Your maneauvering about using intuitions to tweak the priors in a Bayesian framework is clever, but ultimately can’t do what you want it to do. You want our inclinations about odd cases to still have some dispositive value in comparing theories, but that’s only remotely possible if we’ve exhausted pretty much every other way in which those theories are comparable. And it’s going to be exceedingly rare for two real rival theories to be so totally evenly matched that the feather-light weight of such intuitions can do any work in tilting the scales.

    (I also don’t understand why the method, even if it worked, is allowed to stipulate away from worries about how widely shared the intuitions are or aren’t. I, for example, just don’t share your proffered intuition that it is never irrational to believe what you know to be true, though I would agree that it is almost never irrational to believe what you know to be true. Shouldn’t this mean that your argument, even if I granted everything else, should have no traction for me? And if it should, then shouldn’t you care quite a bit as to who does or does not share that intuition?)

    (iii) To apply my general point from (ii) to your argument: in assigning values to the priors of KR and (IT&CE), it seems to me relevant that (IT&CE) has radically more substantial empirical commitments than does KR. At a minimum, (IT&CE) is committed to (a) the existence of a ‘cognitive system’ that can be distinguished from the nervous system more generally; (b) the empirical reality of a function of such a system (whose function is presumably distinct from the function of the nervous system as well); and (c) the empirical reality of an inquisitive/non-inquisitive distinction. Now, (a), (b), and (c) may all be true, though they strike me as currently speculative at best. But, regardless of their actual truth-vaue, they are all serious commitments of (IT&CE) that KR does not share. We should thus significantly down-rate (IT&CE)’s prior, as compared to KR’s, and indeed in a way that is nowhere near compensated by whatever faint intuitive tug might pull in the other direction.

  21. Hi Ram,

    Well, you are missing my first claim, which is that epistemic norms are the ones we follow in determining whether to believe that p. Given transparency, this entails that epistemic norms are norms for determining whether p. Thus it is transparency, or whatever accounts for it, that is responsible for the alethic nature of epistemic norms. If this is right, then we ought to expect that the correct explanation transparency will shed light on how we should understand these norms: whether, for example, we should understand these norms as norms that govern a functional system. It would be surprising if 1) transparency is responsible for the alethic nature of epistemic norms, 2) these norms are to be understood in terms of the function of a system, but 3) the correct explanation of transparency has nothing to do with the function of this system.

    My own explanation of transparency, which is consistent with a purely naturalistic ontology, implies a conception of epistemic norms that is different than the one described in the passage of yours that I quoted, at least as I understand that passage.

  22. Hi Nishi,

    What you’re calling “surprising” — namely the conjunction of your (1), (2), and (3) — is precisely what I’m antecedently inclined to believe. There are lots of properties that epistemic norms have. They are transparent, in your sense. They also govern one of our functional systems. Why should we explain the former in terms of the latter? In any case, as I said, I very much look forward to reading your paper!

    And Jonathan,

    Thanks, as always, for your helpful comments! My responses:

    (i) I think that my current version of naturalism is compatible with the Craig-style approach that you and I both embrace. If you want to know what the contemporary function of a socially shared artifact is (an artifact that arises out of mutual attitudes of some kind, rather than out of a single person’s act of creation or imposition), then consider what sort of hypothetical contract would give rise to such an artifact given our contemporary needs, interests, and circumstances. While our cognitive system is not an artifact, our mutual attitudes can (within limits) impose upon it functions that it doesn’t already have.

    (ii) Intuitions. I do agree with you that intuitions are not good evidence. (In fact, it’s your work that got me to thinking this.) But I also think they have dispositive value, in so far as they set priors. If different people have different intuitions about the same cases, then their priors will be set differently. But that’s not much of a problem, since, in the long run, differences in our priors are swamped by the accumulation of evidence — that’s why it doesn’t matter that much how widely shared our intuitions are. (So long as our priors aren’t set to 0 or 1.)

    I like to think of epistemological intuitions as analogous to the gustatory sensations that one gets when tasting wine. The novice’s intuitions are simple — red or white. But the epistemologist’s intuitions can be much more nuanced and complex — vanilla, a hint of oak, a cherry finish, etc. Of course, even the expert’s intuitions are shown to be wrong if we, so to speak, examine the barrels and find out that there was no vanilla, oak, or cherry anwyhere in the process. So empirical evidence always gets the last word.

    (ii) The three substantial commitments that you correctly spot in IT & CE strike me as so obviously true that they don’t lower the prior probability of IT & CE. But maybe you can tell me why I’m wrong about this…?

  23. Hi Ram,

    I don’t need to disagree with your claim that intuitions can have some dispositive value — all I need is that they don’t have much, and anyway not nearly as much as the various other factors that will come in when we’ve got any two really substantively different theories. (There is a separate issue about how shared they are, in that I still don’t see how your argument is supposed to work at all on someone who, like me in this case, simply fails to have the requisite intuitions.) I am claiming that when we consider (IT&CE) vs. KR, the different commitments of the different theories under consideration will have a far greater impact on their priors than whether any given funky hypothetical case tastes of berries and a hint of apple.

    I offer those three commitments I identified as a particular application of this claim. And note that I don’t need for those commitments to be false, or even to be moderately likely to be false (though see the next paragraph). All that I need is that each commitment has Pr biological difference between inquisitive and non-inquisitive people, of a sort required to ground the attributions of teleology that you have in mind.

    So, given all those concerns, even if they prove to be ultimately addressable, our prior on (IT&CE) should take a significant hit, more than sufficient to swamp whatever adjustments might be due to intuitions. And I think that this generalizes methodologically: if we’re going to pursue this weighing-the-priors approach for theories that we recognize have substantial empirical content, then the place to look to weigh them is in the empirical literature, and not our intuitions. Not (necessarily) because intuitions have no possible dispositive value, but because that dispositive value is teeninsy when compared to that of the empirical commitments of the different theories.

  24. Hi Jonathan,

    OK, so two things:

    (1) Shouldn’t we regard the three substantial empirical commitments of IT & CE as lowering the prior of that view somewhat?

    (2) Why does it not matter if our intuitions about hypothetical cases are unshared?

    I think the answer to (1) is “no”. And I wonder if your inclination to answer (1) in the affirmative isn’t the result of your giving an overly strong reading to the three commitments. For instance, I don’t claim that there is a biological distinction between inquisitive and non-inquisitive creatures. I just claim that there is a distinction — biological or not. (Probably not.) That’s all I need, and that seems entirely innocent to me. Why does it seem entirely innocent to me? At least partly, it’s because I’ve had lots of experience with human beings and with their pets, and I’ve found that at least some of the former (e.g., my colleagues, you, some of our mutual friends) are occasionally somewhat inquisitive.

    The answer to (2) is: it doesn’t matter that much if our priors are different, since, in the long run, evidence will swamp differences in our priors. So if you and I assign different priors to Kornblith’s reliabilism, that’s not so bad: evidence will eventually decide between KR and IT & CE. But even a champion of KR can admit that I’ve put forward an alternative hypothesis — one that squares equally well with the evidence from cognitive ethology — that deserves empirical comparison with KR. It’s that empirical comparison, and not my stuff about priors, that will be dispositive in the end.

  25. With regard to (1), perhaps the real question is what it takes for something to count as a system. I was sorta using biology as a proxy for that question, but it can happen at the level of scientific psychology instead. Here’s a relevant passage from your paper:

    “Suppose we functionally analyze the nervous system at an extremely high level of organization. We might analyze it into such functional components as: perception, memory, reasoning, decision-making, sensation, imagination, and so on. Some of these components can be grouped together under the heading “theoretical cognition”. These are the components dedicated to performing cognition of facts (as opposed to cognition of what to do). It is, of course, an empirical claim that our nervous system has these functional components.”

    The main thrust of my earlier comment should be seen as raising a doubt as to whether or not such a functional decomposition is really possible (current cognitive science seems to indicate that it is unlikely); and, even if it such a decomposition is really possible, whether we can make good sense of attributing teloi to the various pieces of it in anything like the way we do with, say, the circulatory system; and, finally, whether the way that we determine the function of, say, the circulatory system will lead us to attribute different functions to the cognitive systems of inquisitive and non-inquisitive creatures. My counter-argument doesn’t need these to be terribly strong commitments — just empirically real commitments, and potentially problematic ones to boot. That should be more than enough to lower Pr(IT&CE).

    Also, given what you say in regard to (2), I just don’t understand why you bother to go through all that extended rigamarole about the guy who’s blindly following a correct proper linear model for wine, etc. If all you want to do is argue that IT&CE is a candidate empirical hypothesis, then surely none of that wacky stuff is needed. But it seems in the paper like you do want to argue for something stronger — i.e., that IT&CE is in better empirical standing than KR. But then (even pace my main argument about the empirical commitments) the argument will simply fail for those who don’t share your intuition. In which case, you do need to care about how widely shared the intuition is or isn’t. But then you say in the paper that it doesn’t matter to your argument how widely shared the intuition is. So, what gives? Just what is the point of the intuition-mongering in your paper?

  26. Hi Jonathan,

    (1) If contemporary cognitive science can’t give us a functional decomposition of the nervous system that tells us what inquiry is, or what cognition is, then (my money says) that’s because it’s not looking at a sufficiently high level of organization. I would have thought it’s empirically obvious that there’s such a thing as inquiry, and such a thing as cognition, and that these things are (somehow or other) implemented by the nervous system. If there’s evidence against this (and not just a big gap in our knowledge about how high-level functions are implemented), then I’d like to see it.

    (2) I told the story about the wine taster in order to explain why I assign a higher prior to IT & CE than to KR. Whoever shares my intuitions about that case is under some pressure to assign priors the way I do. But you don’t fall into the latter category, so you and I assign different priors, and so the intuition-mongering in my paper will not be of any interest to you, except in understanding why I happen to set my priors as I do. Well, that’s OK, since eventually enough empirical evidence will be accumulated to decide the issue — or so we can reasonably hope!

  27. re: (1) – It’s nowhere near enough, for your purposes, to observe that there is some cognizing and inquiring going on. After all, there’s also nose-picking and checkbook-balancing and opera-singing and voting and all sorts of other activities that we engage in all the time, but for which there is no underlying system. And, moreover, for which there will be no proper function of the sort that evolutionary biology attributes to the circulatory system. So you’re going to have to do a lot more, if you’re going to show that we have a psychologically-decomposable system of pure cognition.

    And the evidence against the existence of a pure cognition system is not just that science merely hasn’t uncovered it yet. Rather, there are a number of theories out there that actively deny that we can parcel off a truth-seeking part of our psychology from the action-guiding part. E.g., the New Look school of a few decades back; or the neo-Gibsonian approach of understanding perception in terms of ‘affordances’ for action. A nice & accessible place to look for an intro to this kind of thinking, under the rubric of ’embodied cognition’, is Andy Clark’s Being There.

  28. Hi Jonathan,

    Excellent point — I shouldn’t have made so much of the nervous system. But whether the system that implements cognition is the nervous system itself, or something bigger (more “embodied”), I don’t see how this can possibly affect my general point in the paper. Let’s consider the very examples that you give.

    Suppose you are taken to an opera for the first time and you see and hear someone singing an aria. You might think “this is ridiculous: what’s the point of this?” Then, if you learn about the history of music, and of opera in particular, you might eventually figure out (on the basis of empirical information, of course) the answer to that question. You might figure out how it is that opera works within a particular musical tradition.

    Again, suppose you have lived all your life in some pre-political society of 50 or so people, cut off from the rest of the world. Then you are taken to a big country and you watch an election. You might wonder “what’s the point of this?” And again, learning about the political system of that country can enable you to answer that question (once again, on the basis of empirical information).

    Same goes for checkbook-balancing. Nose-picking is a more interesting case: a medical friend of mine spculates that its function in toddlers is not to relieve discomfort (which is generated only by our awareness of certain social norms governing personal hygiene), but rather to strengthen our immune system. That, of course, is an empirically testable hypothesis.

    Now, my point is that cognition is exactly like all of the examples that you have listed in just this way: it has an empirically discoverable function. Whether the function is implemented in the nervous system itself, or in something bigger than the nervous system, is beside the point.

  29. (Sorry for the lag — I’ve been travelling.)

    Ram, the ’embodied cognition’ movement isn’t really so much about noting that cognition is implemented in a broader swath of gunk than the nervous system per se. It is much more about exactly that question of how we are best to understand what cognition is for, which will then inform our attempts to understand the actual functional decomposition of cognition. And what the embodied cognition folks claim is that there really is no system of pure cognition in the sense you describe — all cognition is intimately tied-up with the production of action in particular environments, and it is scientifically inaccurate to parse cognition out into a getting-information system and a doing-stuff-in-the-world-based-on-that-information system.

    Now, this is only a potential problem for you if functions have to be at least somewhat part of science. But your last comment seems to leave science behind more or less completely, so maybe this isn’t so. That is, maybe, for your purposes, it’s not a problem if science tells us that there’s no such thing as a pure-cognition system, because from some other, extra-scientific-but-still-empirical perspective, we can usefully talk about such a system. I don’t have any objections to such a perspective (though I would note that your paper, with its talk of the circulatory system, etc., really does suggest that you want a scientific perspective). From such a perspective, perhaps — and I’ll grant it for the sake of argument — we can attribute functions to behaviors themselves, even if such behaviors have no underlying systematic explanation. But you will run into two connected problems, if you go the way you suggest in your last comment.

    First, for almost anything that we want to talk about, there will be a plethora of attributable functions. For most any X worth talking about, we’ll have: (i) the function that we attribute to X to explain X’s origin; (ii) the function that we attribute to X to explain X’s continued existence (which may or may not be the same as (i)); and of course (iii) the explicit purposes adopted by those who participate in and/or deploy X (and obviously there may be a potentially large number of different functions that will arise from (iii)!)

    Considering the case of opera, for example, we might (if we’re feeling particularly Marxist on a given day) explain its origins in terms of amusement for the nobility; its persistence in terms of the bourgeoisie’s interest in taking on the ‘respectable’ trappings of the nobility; and all sorts of different explicit purposes, from of course the aesthetic (musical satisfaction; dramatic satisfaction; some Gestalt of the two) but also to the social (wanting to impress a date; wanting to signal to the voters that one understands their class interests) and to the pecuniary (wanting to make a living as a soprano/composer/conductor/house manager, etc.). If we were to restrict ourselves to a particular kind of scientific attribution of function, then we at least have a hope of getting something like a unique proper function, or at least a fairly constrained set of such functions. But I don’t see how we come even vaguely close to constraining the set of correctly attributable functions, if we go the direction you indicated in your last comment.

    Second, these functions don’t seem to have any normative force whatsoever, beyond the boring conditional sort “if you want to use X to accomplish Y, then you should…” that hardly needs all the fancy trappings of teleology to get going. This worry is connected to the first, obviously, in that if there are so many functions attributable to X, then it is likely that they frequently will pull us in inconsistent directions. Surely the moneymaking and aesthetic functions of opera are not always well-aligned, even if sometimes they are. It’s in exactly the sorts of cases where our various interests in X start to come apart, where we might want some guidance from the philosophy of X — and exactly there that your account will completely break down.

    I should note that something like this problem of normativity would arise on a more scientific-teleology picture as well. As evolutionary psychologists are often at great pains to remind us, just because we evolved some trait to serve some biological purpose, it does not at all follow that it is good or right for us to exercise that trait, sometimes because there is a mismatch between the EEA and the current environment, but also often because doing so would be morally wrong. (E.g., even if you buy the old line about men being ‘designed’ to cheat on their mates, it in no way follows that men should do so.)

  30. Hi Jonathan,

    Once again, you make an excellent point, namely that there are many ways of correctly attributing functions to an activity or institution. I need to make it clear that I am interested in the second thing on your list: attributing functions to explain the CONTINUING EXISTENCE (at present) of the activity or institution in question. I am not interested in attributing functions to explain origin, nor am I interested in consciously adopted purposes (although I spoke misleadingly of consciously adopted purposes in my early reply to Kvanvig above).

    So that is how I want to reply to one of your points. But you make two other points that I should say something about.

    You say that I leave science behind more or less completely. Well, I don’t know what you mean by “science”, so I don’t know whether or not I leave it behind. But I’m interested in empirically discoverable functions. Since I’m not sure what sorts of empirical discovery gets to count as “scientific”, I’m also not sure why it matters whether or not the empirical discovery of such functions is “scientific”.

    Also, you worry that the empirically discoverable functions that explain the continued existence of an activity or institution have no normative force whatsoever. Now there are cases that illustrate this general point. For instance, the continued existence of the Murdoch media empire is explained by the fact that certain functions are well served by Fox News, the New York Post, News of the World, and so on. Nonetheless, those functions have no normative force — at least not for anyone who doesn’t work, or aim to work, for Murdoch. So there’s a case where the functions that explain the continued existence of something don’t have the kind of normative force that I’m after. But I don’t see how the case of cognition could be like this for us. All of us are presumably interested in being good at cognition, no?

  31. One question that suggests itself, but which I’ll set aside for now, is why should that particular function — the ‘persistence function’, we could call it — be the one that we should try to actualize? I’ll set it aside, though, because even granting an answer to it, the account runs into deep difficulties.

    First, determining the persistence function of pretty much anything is a tricky business. To return to our earlier example, it is far from obvious what the persistence function of opera is. It’s not too hard to conjure up a list of candidates, but it’s a matter of serious social history to determine which (if any) of the members of such a list are really the persistence function(s). There are many significant parts of our psychology for which it’s a matter of very open debate what the persistence function is — or indeed whether the part in question even has such a function, or is merely a spandrel. (You might look at some of the evolutionary psychology literature on emotions, and emotional disorders like depression, for an example of the difficulties in thinking through the functionality of a fairly robust component in human psychology.) So it will be a lot harder for epistemologists to make use of your teleological suggestion than your paper seems to indicate — the paper seems to want us to be able to reach a pretty good idea as to the function of cognition while sitting in what we might call a well-read armchair. But in fact it will take a very large amount of hard empirical work to determine what such a function might be, or even whether there is one.

    Our general state of ignorance as to the persistence function(s) of cognition leads us to a second major problem. You ended your last comment by saying, “All of us are presumably interested in being good at cognition, no?” Though I think you meant it as a bit of rhetorical fun, it
    revealed that you take it as obvious that, whatever cognition is, it’s something we should want to be good at. And I want to agree with you about that obvious claim — but the problem is that your account would makes it a very, very non-obvious claim. It’s entirely plausible, for example, that the persistence function of pure cognition (considered sociologically) is the serving of various powerful interests — e.g., that we have an NAS only because we have powerful people who make lots of money off of selling medicines. In that case, it’s not so obvious to me that I want to be good at cognition, if “being good at cognition” means “cognizing in such a way as to make money for various powerful interests”. Or, considered biologically, it’s plausible that pure cognition has no function at all, and is a spandrel of our ‘impure’, action-oriented cognition. In that case, there’d be no such thing as “being good at cognition”! And that’s just for the first two alterante hypotheses of persistence functionality that suggested themselves to me; I’m sure that the real empirical investigation of the matter would have to consider a much broader set of candidates, some of which might seem like something we’d want to be good at, but many of which won’t.

    Now, as I said, I do think it’s obvious that I should want to be good at cognition. But that’s because I know what I want out of cognition — and, moreover, I think I’ve got a line on what we, in our community of investigators, want out of cognition — but what I/we want from cognition might turn out to be something very different from cognition’s persistence function. So it seems to me that the persistence function of cognition that not only would be extremely hard to discover in the first place, but moreover would be of limited normative influence were we to discover it. I’d of course be interested to learn what, if anything, that persistence function is, but that is an interest I have qua philosopher of cognitive science, and not qua epistemologist.

    (As for the question of what I mean by “science”, I mean, well, the stuff that we all unproblematically call “science”, i.e., the stuff done by the people we all unproblematically call “scientists”. I wasn’t trying to make any heavy philosophical hay out of the term, but only appeal to the fact that some scientists have ways of talking about some sorts of functions, but that not all respectable ways of talking about functions have found a home in the professional discourse of scientists.)

  32. Hi Jonathan,

    So you raise two challenges:

    (1) Why do I claim that the persistence function has normative force?

    (2) How can I hope to discover the persistence function, given the general difficulty of discovering such functions?

    Then, in sketching your own view, you say this: “what I/we want from cognition might turn out to be something very different from cognition’s persistence function. So it seems to me that the persistence function of cognition that not only would be extremely hard to discover in the first place, but moreover would be of limited normative influence were we to discover it.”

    I can respond to (1) and (2) above by taking on what you say in the quoted passage. Now, I grant that some of the things that I/we (at least sometimes) want from cognition are very different from cognition’s persistence function. But isn’t there going to be SOME determination of the latter by the former? Why do I/we continue to perform cognition, and continue to make (sadly limited) efforts to improve my/our performance, as opposed to going into a drug-induced stupor? The persistence and development of human cognition is the effect of lots of choices that we make. I would think that there’s going to be some determination relation, however messy, between the ends that guide those choices, on the one hand, and the persistence function of cognition (i.e., the function by virtue of which we continue to perform it), on the other hand.

    So, in thinking about the persistence function of cognition, it’s not bad to start by thinking about the various things that we generally want from cognition. (That’s why I think that the naturalistic line I develop in this paper does not conflict with the Craig-inspired view that both you and I embrace.) That’s not a recipe for discovering the persistence function of cognition — I don’t know any recipe for that (other than something unhelpful like “compare all the hypotheses you can think of against all the evidence that you have)”.

    So why is it the persistence function that has normative force? That functions has (roughly) whatever normative force our choices, or the functions fixed by our choices, have for us who have made those choices. (I don’t know exactly where THAT normative force comes from: why are my choices normative for me? But I assume that there is an answer to that question.)

  33. “The persistence and development of human cognition is the effect of lots of choices that we make.” This is true, but not, I think, in the sense that you need it to be true. Probably for any reasonably complex social phenomenon X, our choices have a big impact on its persistence function — but not, generally, our choices of the form “What do I want X for?” or “What ought we want to do with X?” or anything like that. Rather, for really pretty much any such phenomenon, we take X pretty much as found, and then choose to use it or act within it in various ways according to whatever incentives we find ourselves operating under. But it sure seems to me that we do generally do very little to no teleological deliberation whatsoever about the various social phenomena that structure our lives. Our choices are more about how to maximize the benefits to ourselves and those we care about, given the social landscape that we simply find ourselves in.

    I mean, you’ve got to figure that if anyone is going to think about questions like “what is the purpose of cognition?”, it’s going to be epistemologists — yet it is not a question that has been considered much in the literature yet. So how much should we expect that the rest of the cognizers have made choices about what to do with their cognition? And I for one don’t think I’ve ever asked myself, “Well, do I want to go on cognizing?” — that I’ve got cognition is taken as given, and then I just use it for various ends, the vast majority of which are practical, including the ‘end’ of the intrinsic pleasure of cognition. (Most philosophers, I would bet, rate very high in terms of what psychologists call “need for cognition”: an intrinsic motivation to think about stuff.) Now, maybe I’m just especially unreflective about these issues … but I suspect that in fact I (and anyone else who would take an interest in this blog!) am a few standard deviations above average for the general population in reflecting on the nature, value, etc. of cognition.

    Perhaps a good way to see this point is to consider other social phenomena, such as gender roles, or etiquette (and didn’t Levi-Strauss write a book on table manners?), or voting, or any given national holiday, or … and observe how infinitesimal is the role played by explicit choices about the purposes of such phenomena in their persistence and development. (Voting might be a particularly good example for my purposes, since there is at least some explicit discussion of what we’re supposed to be doing by way of having a social institution of voting — which, it seems to me, has more or less no impact on our actual electoral practices!)

    At worst, explicit teleological choices are epiphenomenal to the persistence of the phenomena; at best, maybe they contribute some very small factor. But that factor is going to be so swamped by the influence of all the other incentives, motivations, and interests involved, that we would be exceedingly irresponsible to rest our judgments of persistence teleology on those choices, when there’s plenty of data out there on the other, much more relevant stuff if only we would go out and get it.

  34. Hi Jonathan,

    I agree entirely that the choices that fix the persistence function of cognition are, by and large, not choices ABOUT the persistence function of cognition. But what follows from that? It doesn’t follow from that looking at those choices won’t help us to figure out what the persistence function is. Looking at those choices might not end up helping us to figure out what the persistence function is, but I don’t see why they’re not as good a place as any to start.

    As I said, I don’t have any instructive recipe for figuring out what the persistence function is, except the uninstructive recipe of “compare all the hypotheses you can think of against all the evidence that you have.” OK, well I put forward one hypothesis, and I also consider a rival hypothesis that might appeal to some reliabilists, and then I argue that my hypothesis handles all the evidence as well as the reliabilist’s hypothesis, and has a higher prior, so it has a higher posterior. (I understand that you don’t accept my argument for the conclusion that my hypothesis has a higher prior. And I think that’s fine, since eventually all our priors will be swamped by the evidence anyway. But at least now we have to look for evidence that will clearly tell against one of the hypothesis, and that seems to me to constitute an advance in the state of the debate.) Well, that adds up to only two hypotheses so far — mine and what I suppose would be a popular competitor — and I’m sure there are lots of other plausible hypotheses that I have never thought of. So once I think of or learn about these rival hypotheses, then we can compare them against the evidence.

  35. OK, well I put forward one hypothesis, and I also consider a rival hypothesis that might appeal to some reliabilists, and then I argue that my hypothesis handles all the evidence as well as the reliabilist’s hypothesis, and has a higher prior, so it has a higher posterior. I don’t believe that this is quite right, as an account of your argument. For the reliabilist hypothesis in question is not a hypothesis about the persistence function of cognition; it is, indeed, a non-teleological hypothesis, and you emphasize in the paper how different you take the ‘KR’ hypothesis to be from your own teleological account. And, because it is non-teleological, our extreme ignorance about the persistence function of cognition does not affect its priors — but it does negatively affect the priors of your hypothesis, which requires a rather particular assignment of teleologies. I think you’ve made a dialectical error in your last comment: you can only downplay the evidential potency of what we can see from our moderately well-read armchairs when you take the priors to be otherwise equivalent, and you’re looking for something (no matter how small) to give the scales a tilt (no matter how minimally). But, in granting that we are in a state of extreme ignorance about persistence functions, you’ve given up the game against Kornblith — as I argued above, the empirical commitments of CE&IT are stronger than those of Kornblith’s KR, and now you’re acknowledging that those commitments are very far from obviously met. So, by the lights of your own framework, and given your recent concessions, you should now significantly favor KR in a pairwise comparison with CE&IT.

    (I understand that you don’t accept my argument for the conclusion that my hypothesis has a higher prior. And I think that’s fine, since eventually all our priors will be swamped by the evidence anyway. But at least now we have to look for evidence that will clearly tell against one of the hypothesis, and that seems to me to constitute an advance in the state of the debate.) I have to disagree, and rather strongly. First of all, it’s a bit … premature, to go casting about for something, anything to give us some preference for one of the two proffered hypotheses when there’s no reason to think that these two hypotheses even come close to exhausting the real competitors (or, indeed, that they are even the main competitors); and, much more importantly, when we recognize that we don’t have even the smallest piece of the evidence that will actually make the difference in determining the truth here.

    Part of what’s so off about this mismatch between the amount of ink & brainpower spilt on the one hand, and the amount of actual effect on determining the truth, is that the paper is put forward as a methodological proposal: here, it says, is how we should go about naturalizing epistemology. But it turns out that what it’s really telling us is how to amuse ourselves by speculating more or less fruitlessly in our armchairs instead of doing the work that would actually, on your accout, provide us with a naturalization of epistemology. To put it another way: if one really believed that (i) the key epistemological facts are those of cognition’s persistence function, and (ii) the real evidence about cognition’s persistence function can only be gained by doing the strenuous empirical work of the sort done by sociologists and the like, then I just don’t see how one can suggest that armchair speculation is even a remotely respectable activity for epistemologists. At a minimum, the paper should have a section where the relevant empirical literatures are canvassed, and then the different hypotheses discoverable in that literature are evaluated in terms of their empirical strengths. Even cooler, of course, would be to go out and do some of the relevant empirical work oneself; I know it’s not part of our training, but if we really came to believe that persistence functions were where it’s at, then it would become imperative for us to pursue such training and/or become much better acquainted with those who already had it. (This, by the way, gives the response to your claim that “looking at those choices might not end up helping us to figure out what the persistence function is, but I don’t see why they’re not as good a place as any to start.” They’re a bad place to start because there are already such palpably better places to start — i.e., the real empirical literature — that they actually seem to distract us from the relevant investigations. They give us an illusory sense of doing something, when in fact we’re merely wasting time that could perfectly well be spent pursuing the evidence that you yourself admit would be much more relevant to the question at hand.)

    Well, that adds up to only two hypotheses so far — mine and what I suppose would be a popular competitor — and I’m sure there are lots of other plausible hypotheses that I have never thought of. So once I think of or learn about these rival hypotheses, then we can compare them against the evidence. I’m pretty sure that, just from my comments so far, we can easily extract several more competitors: the persistence function of cognition is
    (a) serving the interests of the powerful; or maybe
    (b) to guide action in the world; or maybe
    (c) nonexistent – cognition is just a spandrel.

    And here’s one more, off the top of my head, based on some of the work by Jonathan Haidt and others on moral discourse:
    (d) non-rationally influencing the beliefs of others, and defending ourselves from their attempts at the same influence. (I.e., cognition’s persistence function is as a mechanism of persuasion, but not of evidence-based rational deliberation or argumentation.)

    And I produced that short list without even thinking terribly hard for rival hypotheses, and with only a fairly small acquaintance with the relevant research — I’m sure there are others out there in the literature already, if we only looked for them. Not to mention all the relevant evidence already out there in the literature, that we haven’t even started looking at yet! Given that each of these hypotheses has at least some significant empirical evidence in its favor, I see no reason to pay particular attention to IT&CE.

    You know the old joke about the guy looking for his keys under the street light, who later admits that he lost them on the other side of the street but he’s looking for them here because the light is better? Ram, I fear your account is starting to sound like a guy who is sitting in his armchair, casting about for some vague hint of where the keys might be, acknowledging both that they aren’t anywhere near where he sits, and that there’s much, much, much better evidence for their location elsewhere in the house — but who insists, nonetheless, in staying seated right where he is. I have trouble making out such a methodological situation as one that can be recommended with much enthusiasm.

    Which brings me back to a question I started with a few dozen comments ago: why should we do all this teleological stuff instead of just being Craigians? It doesn’t look like you want to get different answers than you would get as a straight-up Craigian, and you’re not looking to make room for evidence that you wouldn’t be able to deploy as a straight-up Craigian. (E.g., if you really wanted to use some evolutionary psychology arguments, I could see how a teleological approach could be a better home for them than a Craigian one.) So, based on how you want to do things, there’s little practical methodological difference between the Craigian and the teleologist. And the Craigian has a much better line on the normativity of her results, since they are based on our reflective judgments of what we should want from knowledge, etc.; whereas the teleologist’s account has, to a good first approximation, no normative value at all — I don’t see how it even begins to give us a reason to offer to someone who would say, “Well, maybe you’re right that X is the persistence function of cognition, but I’m more interested in using cognition for Y.” So, if teleology has nothing to add either methodologically or normatively to Craig, why not just take our Craig neat, and leave the intoxicating-but-ultimately-poisonous bottle of teleology on the shelf?

  36. Neta: OK, well I put forward one hypothesis, and I also consider a rival hypothesis that might appeal to some reliabilists, and then I argue that my hypothesis handles all the evidence as well as the reliabilist’s hypothesis, and has a higher prior, so it has a higher posterior.
    Weinberg: I don’t believe that this is quite right, as an account of your argument. For the reliabilist hypothesis in question is not a hypothesis about the persistence function of cognition; it is, indeed, a non-teleological hypothesis, and you emphasize in the paper how different you take the ‘KR’ hypothesis to be from your own teleological account.

    Neta: The reliabilist hypothesis that I was talking about in the passage that you quoted is not Kornblith’s reliabilist hypothesis. It was the reliabilist hypothesis that I entertain on Kornblith’s behalf in my paper: the hypothesis that true belief is the goal of cognition. That was the reliabilist hypothesis that I was talking about, and it is just as teleological as my hypothesis. In the comment that you quote, I’m talking about comparing two hypotheses about the telos of cognition, not about the nature of knowledge. I also compare two hypotheses about the nature of knowledge — one a teleological hypothesis and the other Kornblith’s. But that wasn’t the comparison that I was talking about in the comment that you quote: I’m sorry I didn’t make that clear.

    Now, you go on to say something that’s relevant to the second comparison:

    Weinberg: And, because it is non-teleological, our extreme ignorance about the persistence function of cognition does not affect its priors — but it does negatively affect the priors of your hypothesis, which requires a rather particular assignment of teleologies. I think you’ve made a dialectical error in your last comment: you can only downplay the evidential potency of what we can see from our moderately well-read armchairs when you take the priors to be otherwise equivalent, and you’re looking for something (no matter how small) to give the scales a tilt (no matter how minimally).

    Neta: I’m not committed to there being any evidential potency to what we can see from our moderately well-read armchairs. (I can perfectly well accept everything that you, Nichols, and Stich have said about the evidentiary impotence of intuitions — indeed, I can go even farther and say they’re not evidence at all.)

    But I am committed to intuition’s being relevant to assigning priors, and you’ve done nothing to argue against that view. Now you and I obviously differ in how important we take it to be that a view respects the conjunction of the requirement of total evidence and the principle that it’s not irrational to believe what you know. I say that the implausibility of any theory incompatible with those two principles gives it a low prior, and there needs to be lots of good evidence to make up for the low prior. I also say that all of the evidence that Kornblith cites is equally probabilified by my own teleological hypothesis.

    Now, I have granted you that my own teleological hypothesis is empirically risky, and has a prior probability that’s much less than 1, and is possibly false. But so what? The question that’s at issue in my argument is this: is the prior probability of my own (admittedly empirically risky) teleological hypothesis higher or lower than KR? And, in light of (what seems to me to be) the very heavy burden that KR has to bear by virtue of being incompatible with RTE & the principle of no irrational knowledge, and the relatively much lighter burden that my own hypothesis has to bear by virtue of claiming that cognition has a normative persistence function, I conclude that KR has a lower prior that my hypothesis, and so has a lower posterior.

    Weinberg: But, in granting that we are in a state of extreme ignorance about persistence functions, you’ve given up the game against Kornblith — as I argued above, the empirical commitments of CE&IT are stronger than those of Kornblith’s KR, and now you’re acknowledging that those commitments are very far from obviously met.

    Neta: Again, so what? We are always in a state of extreme ignorance when assigning priors. That’s just part of what’s involved in assigning priors. On the view I’m proposing, plausibility determines priors, and then we let the evidence rip and see what the posteriors end up being.

    Neta: (I understand that you don’t accept my argument for the conclusion that my hypothesis has a higher prior. And I think that’s fine, since eventually all our priors will be swamped by the evidence anyway. But at least now we have to look for evidence that will clearly tell against one of the hypothesis, and that seems to me to constitute an advance in the state of the debate.)

    Weinberg: I have to disagree, and rather strongly. First of all, it’s a bit ” … premature, to go casting about for something, anything to give us some preference for one of the two proffered hypotheses when there’s no reason to think that these two hypotheses even come close to exhausting the real competitors (or, indeed, that they are even the main competitors); and, much more importantly, when we recognize that we don’t have even the smallest piece of the evidence that will actually make the difference in determining the truth here.

    Neta: Who says that we don’t have even the smallest piece of evidence that will actually make the difference in determining the truth here? We have plenty of evidence concerning the function of cognition (for instance, we know enough to know that all sorts of wild hypotheses are false). But I recognize that it’s an extremely controversial question in philosophy of science how to go about empirically discerning the function of anything, and so I was trying to remain as neutral as I could be on the issue of what evidence is relevant. I don’t want to enter into the controversy concerning how to figure out functions, but we would need to settle that issue before we could agree on which evidence was relevant” …

    Weinberg: Part of what’s so off about this mismatch between the amount of ink & brainpower spilt on the one hand, and the amount of actual effect on determining the truth, is that the paper is put forward as a methodological proposal: here, it says, is how we should go about naturalizing epistemology. But it turns out that what it’s really telling us is how to amuse ourselves by speculating more or less fruitlessly in our armchairs instead of doing the work that would actually, on your accout, provide us with a naturalization of epistemology.

    Neta: I don’t see that. Here’s what I’m saying: you tell me in general how to go about empirically discern the function of an activity, and then I’ll tell you what evidence is relevant to figuring out the function of cognition specifically.

    Weinberg: To put it another way: if one really believed that (i) the key epistemological facts are those of cognition’s persistence function, and (ii) the real evidence about cognition’s persistence function can only be gained by doing the strenuous empirical work of the sort done by sociologists and the like, then I just don’t see how one can suggest that armchair speculation is even a remotely respectable activity for epistemologists.

    Neta: The only value it has is for setting priors.

    Weinberg: At a minimum, the paper should have a section where the relevant empirical literatures are canvassed, and then the different hypotheses discoverable in that literature are evaluated in terms of their empirical strengths.
    Neta: That’s the hoped-for book version. I went 40% over the editor’s budgeted length as it was with this paper.

    Weinberg: Even cooler, of course, would be to go out and do some of the relevant empirical work oneself; I know it’s not part of our training, but if we really came to believe that persistence functions were where it’s at, then it would become imperative for us to pursue such training and/or become much better acquainted with those who already had it.

    Neta: Of course I agree with you about that. (Cooler still would be to manage the publishing company for the book!) So if I persuade anyone that persistence functions are where it’s at, then it’ll be time for some joint empirical work.

    Weinberg: (This, by the way, gives the response to your claim that “looking at those choices might not end up helping us to figure out what the persistence function is, but I don’t see why they’re not as good a place as any to start.” They’re a bad place to start because there are already such palpably better places to start — i.e., the real empirical literature — that they actually seem to distract us from the relevant investigations. They give us an illusory sense of doing something, when in fact we’re merely wasting time that could perfectly well be spent pursuing the evidence that you yourself admit would be much more relevant to the question at hand.)

    Neta: Once again, you tell me in general how to go about empirically discerning functions, and I’ll go and locate the relevant empirical literature. But I’m not going to try to solve the epistemological problems of functions in this paper.

    Neta: Well, that adds up to only two hypotheses so far — mine and what I suppose would be a popular competitor — and I’m sure there are lots of other plausible hypotheses that I have never thought of. So once I think of or learn about these rival hypotheses, then we can compare them against the evidence.

    Weinberg: I’m pretty sure that, just from my comments so far, we can easily extract several more competitors: the persistence function of cognition is
    (a) serving the interests of the powerful; or maybe
    (b) to guide action in the world; or maybe
    (c) nonexistent – cognition is just a spandrel.

    Neta: Good — three more hypotheses! I don’t assign any of them as high a prior as my own, for the simple reason that I don’t find any of them as plausible as my own. But there’s no reason for you to care about what priors I assign. The important point is that all three are empirically testable. Now, precisely how do we go about testing them? What evidence is relevant to their credibility? Good question! — about which I say nothing in my paper. As I’ve said, in order to say something about it, we’d have to solve the epistemological problem concerning functions. You give me the solution, I’ll apply it.

  37. One more thing. Your conclude as follows:

    Why should we do all this teleological stuff instead of just being Craigians? It doesn’t look like you want to get different answers than you would get as a straight-up Craigian, and you’re not looking to make room for evidence that you wouldn’t be able to deploy as a straight-up Craigian.

    Good question, and good point. My response: The whole point of my paper is to say how epistemologists can try to give naturalists what they really want (empirically informed substantively prescriptive epistemology) without accepting a form of reliabilism that a great many epistemologists today would find utterly implausible. Craig doesn’t explicitly do that, but I want to do that in a way that was consistent with his conclusions.

  38. Let me take comment #39 first (as I need to re-read some things in your paper before responding to the rest). I am completely, 100% sympathetic to the goal of showing “how epistemologists can try to give naturalists what they really want (empirically informed substantively prescriptive epistemology) without accepting a form of reliabilism that a great many epistemologists today would find utterly implausible.” But why don’t you think that this can be shown within a Craigian methodological framework? It seems to me that the sorts of considerations you want to appeal to in order to twiddle with priors, could instead serve a real evidential function within the context of a Craig-style investigation. Our reflective consideration of what we want out of cognition plays only an exceedingly indirect role in the teleological framework, but is of immediate consequence within a Craigian framework. And, moreover, the Craigian framework is itself entirely naturalistically acceptable. So, I agree with your goal, but am (as you’ve noticed!) rather skeptical of your proffered means of getting there.

  39. RM: “The reliabilist hypothesis that I was talking about in the passage that you quoted is not Kornblith’s reliabilist hypothesis. It was the reliabilist hypothesis that I entertain on Kornblith’s behalf in my paper: the hypothesis that true belief is the goal of cognition.”

    Ok, that clears things up a lot! But I do seem to have a least a little residual confusion about the paper, though: do you not take yourself to have presented arguments against Kornblith in the paper, then, and just against a teleo-deploying Kornblith*?

    RM: “I’m not committed to there being any evidential potency to what we can see from our moderately well-read armchairs. (I can perfectly well accept everything that you, Nichols, and Stich have said about the evidentiary impotence of intuitions – indeed, I can go even farther and say they’re not evidence at all.) … But I am committed to intuition’s being relevant to assigning priors, and you’ve done nothing to argue against that view.”

    JW: Well, I have been arguing that the differences in priors that can be motivated by these considerations are very, very, very small, which surely should count as an argument against bothering with both the differences in the priors, and the intuitions that are making those itty-bitty differences.

    (Also, once you put it so baldly, I must say I don’t see how it makes sense to grant that something should influence our priors, yet it doesn’t have any positive evidential status. But that’s not a point I want to try to press here, but I reserve the right to press it later, if you try to apply this quasi-Bayesian approach elsewhere….)

    RN: “Once again, you tell me in general how to go about empirically discerning functions, and I’ll go and locate the relevant empirical literature. But I’m not going to try to solve the epistemological problems of functions in this paper.” Now, surely that’s not a dialectical move you can make here. You’re trying to convince us to take up your teleological methodology — but now you’re begging off of the task of actually figuring out how that method would work! It’s like if I said, “here’s my new method in epistemology: for any epistemological question we can formulate, all we need to do is look up the truth in the Great Book of Truths”; then you ask me, reasonably enough, where the book is to be founs; and I say, “oh, that’s the book-location problem — I don’t have a line on that, but I’d be very happy to hear your account of where the Great Book of Truths might be found!”

    Basically, it’s hard to advertise for a new method, if you won’t even sketch how the method should actually work in practice!

    Moreover, even if philosophers haven’t settled all the relevant philosophical questions about functions yet, there’s no reason we couldn’t just take a model from the scientific literature and run with it. Lots of people have offered lots of functional explanations of lots of things, and many of them are, I’m sure, still considered to be in fine explanatory shape, despite the occasional carping of philosophers.

    RN: “Weinberg: At a minimum, the paper should have a section where the relevant empirical literatures are canvassed, and then the different hypotheses discoverable in that literature are evaluated in terms of their empirical strengths.
    Neta: That’s the hoped-for book version. I went 40% over the editor’s budgeted length as it was with this paper.”

    But, Ram, this just underscores my point about how dickering over priors is a distraction from the real work! If you’d cut out all the intuition-mongering, you’d have had plenty of space to discuss some fragment of the scientific literature!

  40. So by my count you bring up four issues:

    (1) Why can’t Craig do everything that I try to do?
    (2) Why do I waste my time trying to argue against Kornblith by assigning priors to his view and to my view on the basis of their relative intuitive plausibility? Isn’t the difference in their priors going to be too small to bother about?
    (3) If I claim that a particular consideration (e.g., intuitive plausibility) SHOULD fix our priors, then am I not thereby committed to regarding that consideration as having evidentiary value?
    (4) How can I advertise a new way to naturalize epistemology, without actually going on to DO it?

    I’ll address them in order:

    (1) Craig himself seems to be more interested in an actual, albeit tacit, contract, than in a hypothetical contract. (Though maybe I’ve misunderstood him.) But I am more interested in hypothetical contracts: what practice of epistemic appraisal WOULD rational, fully-informed parties with our general sorts of ends and circumstances agree to employ? (I suppose you are too, if my recollection serves” … ) So now suppose we’re trying to figure out what practice of epistemic appraisal would be established by means of such a hypothetical contract. Well, we have to take into account the sorts of ends we generally have, and so what sorts of ends the contracting parties would have. So, in order to apply Craig’s method, we need to find out some (empirical) facts about people’s ends. Now of course, people have lots of ends, and a practice of epistemic appraisal needn’t, and probably can’t, be designed to serve all of them. Which ends are relevant? Maybe lots of them, but I’m inclined to think that among the ends that are relevant to designing the practice in question are specifically cognitive ends. Now if I’m right about that, then the application of Craig’s method to the design of a practice of epistemic appraisal requires that we ascertain something about peoples’ ends in general, and something about their cognitive ends in particular.
    (2) You and I obviously differ on how big the difference in priors is between my teleological hypothesis and Kornblith’s Reliabilism. I think my hypothesis is way more plausible, and you disagree. But I think we can just agree to disagree about their relative plausibility, and about the relative priors. I don’t see a problem in our disagreeing about this, so long as we’re prepared to say that evidence can eventually carry the day, and swamp all our priors. I just don’t see that any of the evidence that Kornblith cites in K&PN from cognitive ethology tells in favor of his reliabilism over my teleological hypothesis. So we just need to get more evidence. And I expect that you and I will agree about that.
    (3) I think there’s an important difference between what fixes priors (even: what SHOULD fix priors), and evidence. Maybe I’m using terms like “evidence” in an idiosyncratic way though, so I should offer some clarification. As I use the term, your “evidence” includes all and only what you’re entitled to treat as a fixed data point in inquiry. In the course of inquiry, there are typically lots of things that we’re entitled to treat as fixed data points: for instance, that there really is a laboratory here, that I really did perform the experiment yesterday, that this isn’t all a dream. But I don’t think intuitions fall into that category. There is evidence about who has which intuitions. But to have an intuition is not, on my view, to receive evidence. That’s a very quick sketch of the distinction, but I hope it’s enough for blog-purposes.
    (4) Good question. Of course I can’t give a very good advertisement for a method that’s described so schematically (i.e., “gather evidence to figure out what the cognitive system is for”). The infomercial version of my paper would go as follows: “Do you want a naturalized epistemology, but you’re not willing to be a reliabilist? Then try Craig’s approach — and then see answer (1) above.”

  41. Here’s an odd thing about (1): I agree with every word of it! But nonetheless I just don’t think that it motivates the kind of approach you suggest in the paper. The problem is, the method you suggest would not answer the question, “What are our goals, with regard to cognition?” — which I agree is the question we want to start from — but rather “What ends does cognition happen to serve, such that it explains its continued existence?” My earlier objection (comment #35) about our explicit choices playing potentially little role in the persistence function is very relevant here, in seeing that those two questions are simply not the same. Reading some evolutionary psychology can go a long waay towards helping to see how the two can come pretty radically apart. There are many accounts on offer of the teleology of, e.g., depression, such as depressive episodes signalling to the organism that it has not picked a very good competitive strategy yet & giving it an incentive to pick another. Yet a great many people would find their depression to be deeply non-normative for them, but rather something to be gotten rid of to whatever extent possible, and I see no reason not to agree with them. Our ends, from within our own perspective as agents, need not be aligned with the ends of whatever biological, cultural, or socio-economic systems are actually efficacious in preserving various of our practices and traits.

    Basically, I think your approach flips the relevance of the two questions (about our explicit, reflectively-endorsed ends & about the persistence function). Both are, perhaps, somewhat indirect evidence about the other. But our interest as Craigians is primarily in the former, and thus our interest in the latter is indirect at best. Your account, however, as offered in the paper, locates our primary interest in the latter, and the former has the status of merely very indirect evidence. This is why you end up with the weird consequence that it becomes a real live epistemic possibility that our cognitive ends are, e.g., serving the wealthy & powerful. But that just shouldn’t be a real normative option for us. (This is of course relevant to (4) as well — it’s not a good advertisement for a method if you seem to open up options that we just don’t want to have open, without showing how we have any reason to expect that they’ll be closed off again.)

    As for (2), I’m still confused as to the relationship between the CE&IT line and Kornblith’s actual form of reliabilism, which is not robustly committed as to the teleology of, or even the existence of, pure cognition. I had been thinking all along that the argument was meant to be between those two, in which case my worries about the different empirical commitments come to the fore. But you said, a couple of comments ago, that you were only considering a teleological version of Kornblith. So let me ask my question again: do you take yourself to be arguing against Kornblith himself, or with his teleologically-committed counterfactual twin?

    As for (3), you’ll need some other term for stuff that is evidence-ish, i.e., the stuff that can rationally influence our beliefs, even if it’s not evidence per se. Maybe we can use “positive evidential status”, which something can have some of without actually qualifying as “evidence”. It seems now that you’re committed to intuitions having a fair amount of positive evidential status, even if they aren’t evidence per se. If so, then I don’t think you can just sail around the Weinberg, Nichols, & Stich-type arguments by saying, “Oh, well I don’t think intuitions are evidence“, since (if I may speak without consulting my co-authors) we take ourselves to be offering an argument against attributing much, or even any, positive evidential status to intuitions. But perhaps we can set that particular set of issues aside here. I think it’s still an objection to your approach to say that the stuff you’re appealing to isn’t evidence, but there’s lots of evidence out there — and given a choice between appealing to evidence and appealing to stuff that’s not evidence, shouldn’t we skip the not-evidence stuff and cut straight to the evidence?

  42. (1) You distinguish two questions:

    “What are our goals, with regard to cognition?”

    “What ends does cognition happen to serve, such that it explains its continued existence?”

    Now I take it that the second question is supposed to be about what we’ve been calling persistence functions, so I would rephrase it as

    “What ends does cognition FUNCTION to serve, such that its so functioning explains its continued existence?”

    Now that I’ve rephrased the question that way, I’m wondering to what extent it differs from the first of your two questions. I think that part of the answer to the first of your two questions is: we aim to have true beliefs that we can defend against reasonable challenges. But isn’t it part of what explains our continued engagement in the practice of inquiry this: it functions to give us true beliefs that we can defend against reasonable challenges?

    Again, I’m thinking of the cognitive system as something like a naturally occuring functional system that sufficiently sophisticated creatures can convert into a tool, in order to serve one or another function. The persistence of that tool can then be understood in terms of its serving that function, even if the persistence of the cognitive system in non-sophisticated creatures cannot be understood in terms of ITS serving that function. (Imagine a medically advanced race of creatures who decide to rearrange their circulatory systems on the outside of their bodies, to serve as decoration, just as the Pompidou Center in Paris has its plumbing on the outside for decoration. Then these people would have given the circulatory system a new function — a function that it didn’t have before.) So I’m afraid I’m still missing something here, about how the two questions above are supposed to come apart.

    (2) I take myself to be arguing both against Kornblith’s Reliabilism (in most of the second half of my paper), but I also argue (in a brief section in section V, I believe) against a teleological hypothesis that I thought Kornblith might be friendly to. So the answer is: both, in different ways.

    (3) I need to understand how the findings that you and S and N have brought to light could be used to argue against the use of intuitions to establish prior probabilities. Intuitions are culturally relative — well and good. So too are prior probabilities. We all start from what seems plausible to us, and of course that’s going to differ for different people, no? What matters is that the differences get washed out by evidence sooner or later. What am I missing?

  43. Regarding (1) – It’s of course possible, under very special circumstances, for us to take such complete control of some aspect of our lives that the answers to the two questions converge. But that’s pretty clearly not the case with most complicated social aspects of our lives — i.e., almost all of the political and economic and cultural aspects — and I have yet to see any evidence to suggest that it’s particularly true of the epistemic, either. Here’s a different way of thinking of the two questions: “If we had total control over the practices of cognition, what ends ought we sculpt those practices to serve?” vs. “What function does cognition happen to function to serve, for whatever it is that controls the practices of cognition (perhaps various evolutionary & cultural forces, etc.)?” Only when the “whatever” of the second question is us, can the two questions converge.

    (Btw, I consider this question about how best to take the lesson from Craig that we both want to take, to be the most interesting part of this exchange — which is not to say that I’m not finding it fun & interesting at all points!)

    regarding (2) – Ok, but then I still don’t see how you have a functioning argument against Kornblith. I don’t think you’ve really addressed the fundamental objection that your teleo view has significant empirical commitments that Kornblith just doesn’t have, and thus your theory must have its priors downgraded accordingly. (I hope by this point in the exchange you will agree with me that we just can’t be entitled to any claims that it’s obvious that those commitments are met.) For a robustly empirical issue like the actual persistence function of cognition, surely such considerations will far outweigh any influence intuitions can legitimately have over priors.

    Part of the reason that the empirical commitments must weigh much more heavily than issues about the principle of no irrational knowledge, is that the probability of the principle of no irrational knowledge itself will get significantly reduced if we conditionalize on the teleological view — because, on that view, the principle now becomes hostage to the empirical vagaries of the persistence function of cognition. We might have, pre-teleo, taken ourselves to have some sort of reasonably apriori access to things like that principle. But, if we go post-teleo, we have to acknowledge that we just have no good reason to endorse such principles: they are part of the theory of knowledge, which is to be derived from an account of the persistence function, which is a wholly empirical affair, about which we still don’t know much. On the teleo account, we should indeed treat any intuitions we have about knowledge with great skepticism. (This is totally independent of Weinberg, Nichols, & Stich-type worries about intuitions) Thus, KR’s inconsistency with that principle should pose much less of an objection to KR on your teleometaepistemology, than it would on other, less radically empiricalized metaepistemologies.

    regarding (3) – I have been taking it all along that we are not interested in whether intuitions, as a matter of empirical fact, happen to play a role in setting our subjective priors. I have thought, rather, that the question was one of whether they may play a rational role in the setting of such priors. If that’s right, then I don’t see the force of the anthropological claim that prior probabilities are culturally relative. But anyway, I guess I don’t need anything so strong as that the intuitions about bizarre philosophical cases can play no role in setting our priors — maybe even horoscopes and fortune cookies can play legitimately some role in setting our priors — all I need is that the WNS data (and the Machery et al. data) significantly undercut the degree to which such intuitions can play such a role. If that’s so, then I get more than enough to make my case, along a couple of independent lines: the intuition-nudge to the priors will clearly be outweighed by the empirical-commitments nudge; and moreover it will be even clearer that we shouldn’t be dickering over such small differences priors when there’s actual evidence out there to be considered.

  44. (1) What do you mean by “total control” in your response to (1)? Do I ever have “total control” over my actions, in your sense of the phrase? Of course, my action is in some sense constrained by the laws of nature, by features of my body and the Earth, and countless other things. Is this incompatible with my having “total control” in the sense you intend? I’m inclined to think that all that’s necessary for the two questions to coincide is total responsibility. Whether or not cognition is in our control, we are responsible for it, and that’s enough for my purposes, or so I’m inclined to think. Is there a problem with that view?

    (2) Of course my theory has empirical commitments that KR doesn’t, but I still regard those commitments as very plausible, and so as not downgrading the prior of my thesis nearly as the considerations I raise against KR downgrade its prior. You, of course, disagree with me about what is more plausible than what, and so naturally you disagree with me about what has a higher prior than what. And this difference between us brings us to the next issue…

    (3) You and I can agree that plausibility fixes priors. But you ask why it OUGHT to fix priors. And the best reply that I can give to this is a rhetorical question: given that we’re talking about fixing priors, and so we have to rule out evidence as irrelevant, is it more reasonable to fix your priors by what seems plausible to you, or by what doesn’t seem plausible to you? The WNS data shows that, by fixing your priors this way, different people will have very different priors. And that in turn seems to tell against the reasonableness of placing much weight on priors. I’m happy to agree with all of that, but given very weak assumptions that I make about evidential support, we do have to place SOME weight on priors. That bit of weight is doing some of my argumentative work. And the rest of that work is done by the fact that the evidence from cognitive ethology and from Ameliorative Psychology is just as likely to obtain if my teleological hypothesis is true, as it is if Kornblith’s reliabilism is true, or Bishop and Trout’s SR is true.

  45. re: (1) – I’ve been mulling it over, and I can’t see how responsibility is an appropriate concept to bring in here, in a discussion of persistence functions. Clearly responsibility isn’t necessary for a persistence function, since evolution provides a great many persistence functions, but (pace Intelligent Design supporters) there’s simply no one there to count as responsible for it. The same holds for most social structures as well — they emerge out of complex interacting social forces, but there’s no one to whom responsibility could be appropriately attributed. And responsibility isn’t sufficient, either: suppose it is my job to be an engineer helping design some device, in which case I might really be responsible for its persistence function; but then suppose (as is actually the case) that I am an absolutely terrible engineer, and am unable to fulfill this responsibility, and so the device gets made without my input. In this case, I had a responsibility, but ultimately did not contribute to the persistence function.

    Anyway, it’s seems obvious that there’s no one to whom we can attribute responsibility for cognition. We can each of us be held responsible for what we do with cognition as we find it, but it doesn’t even begin to follow that we are responsible for its persistence; in much the same way we can be held responsible for what we do with our arms and legs, but our reflective choices we make now about limb usage are mostly irrelevant to determining their persistence function (though over a very, very, very long time, those choices might have some small impact on that function).

    re: (2) & (3) – What’s becoming clear is that you want the assignment of priors to be two things at once that they just can’t be. On the one hand, you want them to be fundamentally subjective (you assign yours your way, and I’ll assign mine my way, and neither can gainsay the other; you also need this in order to avoid my argument above that you should update your assignments of priors to reflect that the teleological framework implies that our intuitions about knowledge should be taken as nigh-well worthless) and pre-evidential (“…so we have to rule out evidence as irrelevant….”). On the other, they’re supposed to be something that it makes sense to evaluate as rational or not, and also something that it makes sense for us to offer up publicly to each other, as a contribution to the epistemological project. The tension between these two strands of your ‘priorist’ method is, I think, obvious. If priors are so subjective, then they are not worth our pestering each other with; and if priors are the sort of thing that can be evaluated as rational, then considerations of evidence can & should come into play.

    I think the right way to resolve this tension would be to give up on the model of subjectivist, nonevidential priors as something to appeal to in philosophical debate. This model seems to take “prior” to mean “prior to all evidence whatsoever“. I’m not sure that that really makes any sense, and at a minimum I don’t see how by this point in our lives we are in such a position with regard to questions of what cognition is & what it’s good for — we’ve been exposed to far too much evidence already. Now, there’s a perfectly respectable notion of “prior” that doesn’t have all these problems, namely, “prior to some specific piece of evidence” — a sort of summation of where one’s evidence to date has left one with regard to whatever question is at hand, such that one can now decide, upon receiving a new piece of evidence, how to adjust one’s assignment of probabilities to the various possible answers to that question. If I take a test for disease X with a small-but-nontrivial rate of false positives, and it does (eek) come out positive, then we can reasonably ask, what was the prior probability that Jonathan had X? If there’s a low enough incidence of X in the general population, and I don’t have any known risk factors, my priors might be low enough that my final probability might still be quite low (whew). But note that in cases like this the assignment of priors is very much based on evidence — evidence about the demographics of the disease, evidence about my background and medical history, and so on.

    I’m not saying that a subjectivist notion of priors is incoherent on its own, btw, just that it doesn’t make sense to report on one’s priors if that’s the kind of priors one has in mind; nor does it make sense to speak of “putting weight on” such priors, as if they were some special sort of evidence-but-not-really-evidence.

    So I’m inclined to dissent from this characterization of the exchange: “You, of course, disagree with me about what is more plausible than what, and so naturally you disagree with me about what has a higher prior than what.” Now that I see what kind of priors you seem to want to talk about, I must confess that I just don’t know what my priors were, and I don’t much care about what they might have been (it was surely a long time ago that I might have had them). I take myself to have offered some evidence about the likelihood of various empirical components of the account, not merely a subjective report of my own non-evidence. That the apparent method of the paper seems to license a kind of ‘his priors/her priors’ mode of discourse, in which opponents’ contentions can apply no rational force upon each others’ deliberations (because it’s all subjective), is a good sign that this is not a method we should want to make general use of epistemology. Intuitions are themselves damnably subjective, but at least they are aiming for objectivity, and thus expose themselves to critique. Subjective priors don’t even do that; which is why they ought never be appealed to in public discussion.

  46. Hi Jonathan,

    Re (1), you say “The same holds for most social structures as well — they emerge out of complex interacting social forces, but there’s no one to whom responsibility could be appropriately attributed.” I agree with the first conjunct but not with the second. To the extent that I am complicit in the persistence of the social structure, I share responsibility for its persistence. Or so it seems to me. Case in point: to the extent that I am complicit in the persistence of a particular activity (e.g., inquiry), I share responsibility for its persistence.

    More interesting to me though, is the issue you bring up concerning (2). Priors are fundamentally subjective, but it doesn’t follow that they’re not rationally constrained, and it also doesn’t follow that they’re not rationally constraining.

    How are priors rationally constrained? If, for example, I assign priors that are too high or too low, then I will be needlessly slow in learning from evidence. (Indeed, if I assign priors of 0 to true hypotheses or 1 to false hypotheses, then conditionalizing on my evidence will never correct me.) Indeed, much of what you and I have been arguing about is whether or not my relative priors for the various theories I consider are irrational. Note that I did not say that my priors were not rationally constrained — I said only that there was nothing irrational about my assignment of priors. And that’s compatible with there being nothing irrational about your assignment of priors either. Rational constraint does not have to single out priors uniquely.

    How are priors rationally constraining? By entering into the mathematical determination of posteriors. I don’t know of any account of empirical support that manages to get priors out of the picture altogether.

    To say that priors are pre-evidential is of course not to say that they are assigned by some mythical tabula rasa agent. Someone who assigns priors has, of course, lots of background beliefs and experiences, and these beliefs and experiences will causally affect her judgments of what’s plausible or not, and so (on my account) causally affect her priors. But for these beliefs and experiences to causally affect her priors is one thing, for them to count as evidence is another. (They may do both, of course, but then they are doing two distinct jobs, and we can consider their performances of those two jobs one at a time.)

  47. Ram,

    It is very interesting that you say this,
    “To the extent that I am complicit in the persistence of the social structure, I share responsibility for its persistence. Or so it seems to me. Case in point: to the extent that I am complicit in the persistence of a particular activity (e.g., inquiry), I share responsibility for its persistence.”
    I wonder if that is true. You are certainly not causally responsible, or even complicit in the cause of, there existing an institution or social structure or activity of inquiry. There would no doubt remain an activity of inquiry should you not participate. There would not doubt have been such an activity had you never participated. Certainly the participation of any particular inquirer is redundant to the persistence of inquiry. It is hard to see how you are responsible in any way for it. I understand the pull to attribute at least some responsibility for it’s existence to each of the participants. But this model can be completely mistaken. Pushing a slow-moving train does not afford you some responsibility for its forward motion. You’re contributing alright, but your contribution is completely inefficacious.

  48. Hi Mike,

    Hmmm… You sum up by describing my role in the train case as follows: “You’re contributing alright, but your contribution is completely inefficacious.” What sort of “contribution” am I making, if not a causal contribution? I grant that the train would be moving forward without me, but am I not part of the cause, and so partly causally responsible? When 1000 people conspire to murder someone, aren’t all 1000 of them responsible, even if each one was redundant given the others?

  49. Ram,
    I doubt that responsibility is divided evenly among equal contributors, if that’s what you’re suggesting. Here’s an example similar to one from Don Regan. Suppose you have a choice between (i) helping two people save six lives or (ii) saving one life on your own. In case (i) perhaps you have to lever the six people to safety by standing on some platform with the other two people. In case (ii) perhaps you have to wade into some water save one person. You can’t do both. Now suppose that were you not to stand on the platform the weight of the other two people would be enough to lever the six people to safety. On the account of responsibility you seem to urge, you ought to get on the platform anyway. In that case, you are responsible for saving two lives. If you choose to save one person, you’re responsible for saving only one. But certainly that cannot be the right answer. So responsibility cannot be distributed evenly among equal contributors.

  50. I’m pretty sure that Mike is on the right track here. But even in the cases under consideration, there’s a clear enough story to be told about the way in which the individual’s choices and actions can influence the course of events. With a vast, highly-distributed, and deeply-entrenched social practice, however, there may be no such story to be told. In such matters, we are less like people pushing the train and more like passengers on it.

    Consider, for example, a social phenomenon that is tragically of moment this week: the location of our cities. What responsibility do we have in their maintaining their current locations? Perhaps we will face an actual decision about the location of NO in the future, but put that aside. Surely if everyone got together and made certain sorts of decisions, we could all up and relocate Bloomington or Chapel Hill, or even Boston or Washington. Does this mean that we are responsible for those cities’ current location? And, more to the point, are their continued static locations to be explained even slightly in terms of anyone’s reflective judgment as to where it would be best for those cities to be? (Clearly not — no one thinks that Washington is located in a very good place, especially if they’ve ever spent a summer there!) Similarly with the complexly interacting fabric of our psychological hardware & rich social practices that comprise cognition. Maybe if we all wanted to, we could change some aspects of it. But I doubt it; and even if we could, it doesn’t follow that our choices about what to do with it play any nontrivial role in determining its persistence function.

    As for the stuff on priors, I’m just not seeing how any of your last comment on it was relevant to my objection that priors, in your construal, just aren’t the sorts of thing that can be rationally constraining on or constrained by other people. And that, therefore, they are just the wrong sort of thing to talk about in public, and especially to cite in public argument. Nothing that about intrasubjective rational constraint seems to bear on that argument.

    Also, this is incorrect, as an account of what I have been arguing about anyway: “Indeed, much of what you and I have been arguing about is whether or not my relative priors for the various theories I consider are irrational.” I’ve been trying to argue about what our evidence is for various claims, and to argue in particular that the current state of the evidence indicates that your hypotheses about the persistence function of cognition are much much likely to be wrong than are Kornblith’s non-teleological claims. Part of this argument is that one piece of your evidential picture — the intuitions about the funny cases — should be, on your own overall account, very seriously downgraded. But that’s really only applicable if you take such intuitions to be evidential at all. If you don’t, then my argument is very easy: the actual evidence points completely in Kornblith’s favor. (At least when comparing only KR and CE&IT, that is.) Your main move in trying to avoid this result seems to be to re-categorize my appeal to evidence as a mere bit of autobiography of my own priors. And that’s what I’ve been trying to resist in the last few exchanges.

  51. Hey Mike and Jonathan,

    Let’s distinguish two claims:

    (A) If x is partly causally efficacious in producing outcome Z, then x is partly responsible for Z’s obtaining.

    (B) If x and y are equally causally efficacious in producing outcome Z, then x and y are equally responsible for Z’s obtaining.

    I agree with you guys that (B) is false. But I’m not committed to (B) by virtue of maintaining (A). And I do maintain (A). Now about Jonathan’s case of NO: I think all American citizens are responsible for having allowed our elected federal representatives to refuse to provide the funding for levee maintenance that was requested months ago by New Orleans officials. Not that it was then irrational for us to allow this request to be refused, but having allowed it, we are all responsible for what we now see. (Just as we’re all responsible for every Iraqi civilian who’s killed with our tax dollars.) I don’t see why the location of our cities is any different. The convenient thing about big bureaucracies is that they can diffuse responsibility so widely that people can make themselves feel as if they’re not responsible for all sorts of calamities.

    Now about priors: I take back my comment regarding the issue of whether we’ve been arguing about the rationality of my priors. But can priors be rationally constrained by other people? Well, they can be rationally constrained, and other people can point out the constraints on them. Isn’t that all I need for it to be reasonable for people to discuss whether or not a particular set of priors are irrational?

  52. Hey Ram,

    It’s picking nits (I’m not suggesting that this is news to you at all), but it is not quite (A) that you want to defend. You don’t want causal responsibility (full or partial) to entail (moral) responsibility (full or partial). No doubt your dog is causally responsible for lots of things that you don’t want to hold it responsible for in any other sense (but, if not your dog, then certainly mine). You want causal responsibility + whatever else is required for moral responsibility to yield (moral) responsibility. The right move, for your position, is to simply beg off any view on the way to assign degrees of responsiblity. And that is what you effectively do. So nothing’s especially lost in not committing to (B). But reconsider the train analogy. It’s not true in that case that your contribution makes at least some difference, though you are contributing all the energy you can. Obviously the energy that you transfer to the train-car is absorbed long before it can affect the motion of the train. The energy is never transferred to the wheels since it is dissipated long before that. Jonathan might have something similar in mind. In many complex social contexts, contributions dissipate long before they make ANY difference at all to the persistence of any practice or social structure.

  53. Hey Mike,

    Of course you’re right: I don’t want to say that beasts or babies are morally responsible. I also don’t want to say that S is responsible for an outcome that S’s actions were not causally efficacious in producing. Of course it’s a very tricky issue (or bunch of tricky issues) when someone’s actions are causally efficacious in producing some outcome. I hope I don’t need to figure out all of that right now. What I need is just that the persistence of inquiry is the result of a whole bunch of particular actions of inquirers. And that still seems right to me.

  54. Let us recall that the matter at hand concerned the following two questions:
    “What are our goals, with regard to cognition?”
    “What ends does cognition function to serve, such that its so functioning explains its continued existence?”
    The former is the question of the Craig-style naturalist; the latter, the question of the Neta-style naturalist. And I was trying to argue that we should prefer to approach epistemology by means of the former, to which Ram had replied in part by suggesting that the two questions would converge, should appropriate conditions obtain. The debate, then, concerns what those conditions are, under which the Craig-question and the Neta-question will converge.

    So Ram’s suggestion now appears to be that those questions will converge if we are responsible for cognition; where it is enough for responsibility that cognition persist due to the actions of cognizers. Let us call this ‘minimal responsibility’, so as to bracket any question about its connection to our ordinary practices of responsibility attribution (which I think it’s pretty far removed from; but I don’t think it was really meant as a substantive descriptive thesis anyway, so we can just put aside the issue of whether it is a part of our practices).

    Now that we have the suggestion formulated, it becomes clear what we need to do to find counterexamples: look for any phenomenon X such that (a) we are, in this minimal sense, responsible for X; but (b) the Craig-question and the Neta-question receive different answers. And I think that, if we look at almost any reasonably complex socially-produced system or institution, it will serve as an appropriate counterexample.

    The location of cities will serve as a counterexample, for starters. If we ask, “what are our goals, regarding the city of Washington?”, we ask what we want to do with the city, what we might make of its future — perhaps we should make it into a shining example of a well-run American city? Maybe we shouldn’t bother having any goals for it at all, beyond providing sufficient security for the well-functioning of the federal government? Maybe it should be allowed to set its own goals, with ‘home rule’? But if we ask, “What ends does Washington function to serve, such that its so functioning explains its continued existence?” then clearly none of those answers just canvassed are even remotely possible answers.

    Or consider X = the current state of our voting system. The answer to the Neta-style question is something like, “to serve the powerful interests who want to discourage voting by the underclasses as much as possible, and to undercount what votes they do cast as much as possible”. I doubt anyone here would offer anything like that as the answer to the Craig-style question about the voting system.

    Or consider X = prescriptive grammar rules. One Craig-style answer would be the promotion of a standardized system of communication; but most nit-picky rules (like those against split infinitives, etc.) only function as a form of status enforcer and social marker.

    Note that you don’t have to agree with any of my particular suggested answers. The important thing is only that the range of possible answers, for the two teleological questions, are divergent in cases like these. Yet they are all cases in which we have minimal responsibility for the phenomenon. So minimal responsibility, it seems, is just not a sufficient condition for the convergence of the two teleological questions.

  55. I agree with every word of Jonathan’s first three paragraphs (except for the parenthetical material concerning responsibility). I also agree that Jonathan’s examples, provided in the following three paragraphs, are examples in which the answers to the two types of questions that he distinguishes in paragraph 1 diverge. They diverge, as Jonathan points out, because we cannot the persistence of the social entity in question is not explained by our realizing our goals with respect to that social entity.

    Now, it is, of course, an empirical question whether or not cognition is like those examples. If it isn’t, then my approach and Craig’s approach will converge. If it is then our approaches diverge. And if our approaches diverge — if in fact, the persistence of cognition is not explained by our own goals with respect to it — then, I say, knowing is not a matter of realizing our own goals with respect to cognition, but rather a matter of realizing the persistence function of cognition (cognition’s “own goals”, so to speak, rather than our goals with respect to it).

  56. Quick follow up on my last email. In the event that our goals and cognition’s “own goals” (i.e., its persistence function) differ, why should knowing be identified with the latter and not the former? Consider the norms or ideals that apply in the examples that Jonathan gives in his last email. For instance, is grammaticality a matter of following the rules of our actual language, or following the rules of some language that is ideally designed to satisfy our own goals? Is good urban planning a matter of doing the best we can with current urban material, or rather of doing the best we can abstracting away from the limitations of current urban material? The pattern seems clear.

  57. I don’t think that those cases make for a very apt comparison between the two ways of thinking about the goals of/for cognition. Descriptive grammaticality just isn’t the right kind of thing to look to here, since it has nothing to do with goals — one can think, as Chomsky does, that language has no persistence function whatsoever, but still be committed to all sorts of facts about what is or isn’t, as a matter of fact, grammatical in a given language. That’s why my earlier example was in terms of prescriptive grammaticality rules, things like no-split-infinitives or don’t-end-sentences-with-prepositions. We agree that the question when evaluating any such rule is, “does this rule serve the function of cognition?” We disagree, however, on what it means to answer that question. For you, it should be fine so long as the rule serves whatever the persistence function of such rules are — and if that function is to enforce class boundaries, and the rule does a good job of enforcing such boundaries, then prima facie it’s a rule we should follow. For me, it will be fine if it serves whatever function we would reflectively choose for such rules — namely, the promotion of clear communication, which these rules tend not to promote.

    Also, I don’t see why you think that issues about idealized vs. realistic ways of achieving the goals are at stake here. We’re arguing about what fixes the goals; it is an orthogonal question how best we should try to promote those goals, once we decide what they are. So when you ask the question, “Is good urban planning a matter of doing the best we can with current urban material, or rather of doing the best we can abstracting away from the limitations of current urban material?” I heartily agree with you that the former, not the latter, is exactly what we should want to do. The question we’re arguing about, though, is how to understand what “the best” means here — best, by the lights of the persistence function of traffic planning? Or best, by the lights of our own goals for traffic planning? (Since traffic planning is something that we do have fairly direct control over, I do think that it is a case that might have more convergence between those two questions, by contrast with things like cognition.) The goals may be idealized (though they don’t have to be); but I agree that the means should not be.

    With the differences between the two views clearly on the table, here are three reasons why I think we should prefer the “our goals” view over the “persistence function” view:

    (1) Normativity – There’s no residual question as to why we a reason to follow rules that promote our own goals. But there’s no particular reason why we should follow rules that promote goals other than our own, including the goals produced by the persistence function of a system.

    (2) Metaepistemological skepticism – On both views, what our epistemic norms are is a contingent & empirical matter. But on the “persistence function” view, it is not just a contingent & empirical question — it is a highly speculative contingent & empirical question. As the discussion so far has shown, we’re pretty deeply in the dark as to what the persistence function of cognition is. So we should, on that view, take ourselves to be more or less completely ignorant as to the right norms of reasoning, until we learn a great deal more about the structures in which cognition is embedded.

    (3) Odious possibilities – As perhaps a corollary to (2), the “persistence function” view leaves open the possibility that the right thing for us to do, by way of cognition, is to serve the interests of the powerful; or to non-rationally attempt to compel the assent of others; or maybe there just is no such thing as ‘the right thing for us to do, by way of cognition’, if in fact cognition just lacks a persistence function. If we are choosing a methodology for epistemology, a method that leaves such options open is perhaps not a method that we should look to embrace. The “our function” view does not leave those options open, since we can easily rule out such icky ‘goals’ for cognition, merely by reflectively preferring against them.

  58. Two possibilities. Either:

    (A) We discover that cognition serves our purposes (whether or not serving those purposes has been its persistence function), and we continue to engage in cognition as we did before, and just because we’ve found that it serves our purposes, or

    (B) Not.

    If (A) obtains, then the persistence function of cognition switches to being our own goals: we have now become responsible for its persistence, and for our own goals.

    If (B) obtains, then why are norms of cognition any more “normy” than norms of digestion or circulation? And if they’re not, then (1), (2), and (3) are not problems for my account: they’re precisely what we should expect, no?

  59. “If (A) obtains, then the persistence function of cognition switches to being our own goals: we have now become responsible for its persistence, and for our own goals.” I still see no reason to expect this to be true, for the reasons given a few comments ago: the notion of responsible that can apply here is far too weak to bear on the persistence function.

    Your (A) and (B) are best thought of as ends of a spectrum, not as a binarism. To the extent that (A) is true — i.e., where our goals and the persistence functions overlap — it’s something of a lucky break for us. But the persistence function would still be normatively and methodologically superfluous in those conditions. Those norms have force for us because of our goals, not because of the persistence functions; and it will be far more direct to learn of those norms by considering the goals, not by seeking out the persistence functions.

    To the extent that (B) is true — i.e., where our goals come apart from the persistence function — then the ‘norms of cognition’ in the persistence function sense are indeed normatively inert for us. But our norms for cognition are normatively compelling upon us, in that they further our goals.

    One kind of situation where the nature of this divergence becomes clear, is where we are considering a new rule (or changing an old one). We would be debating what sorts of considerations we ought refer to, in making such a decision. Whether that rule serves the persistence function? I don’t see why I should particularly care if it does or doesn’t. But: whether the rule serves my own goals? Of course that will be a legitimate & pertinent kind of consideration.

  60. My wording of (A) and (B) was intended to make them logically exhaustive. Jonathan and I disagree about how likely it is that (A) is true, but never mind that.

    The point I want to insist on is this: if (B) is true, that doesn’t make persistence function normatively inert for us. Functional systems like circulation and respiration are not normatively inert, even though their normativity has nothing to do with our desires, goals, or ends. There’s such a thing as good circulation, or good respiration, and having these things is intrinsically good (even if in some cases they’re instrumentally bad, because, say, they end up exacerbating some problem that we have, or otherwise frustrating our goals, or we just happen not to want them). What goes for circulation and respiration goes for cognition: the goodness is intrinsic, and we have a reason to want them, though this reason is independent of all of our ends and desires and goals. Jonathan may disagree with me about natural intrinsic normativity, but then we have a big metaphysical dispute on our hands.

    As for what we debate when considering a new epistemic rule: I expect we’d debate lots of things, like how easy it is to implement the rule on the kind of machinery that we can afford given our present budget. In general, I think Bishop and Trout are right about what we’d debate. But, as I argued in the paper, there’s no straightforward inference from what we’d debate to what makes a rule epistemically good.

  61. “But, as I argued in the paper, there’s no straightforward inference from what we’d debate to what makes a rule epistemically good.” I just gave the paper a (much too) quick re-read, and I didn’t see the argument in question. Could you point me in its direction?

    Your assertion in this context that the proper functioning of various systems is intrinsically normative reveals a confusion that is not uncommon these days: a confusion between what is normative in some selective or persistence sense and what is normative for us. (For evidence of how common this confusion is, see some of the controversy that has attended my brilliant colleague Lisa Lloyd’s recent book, The Case of the Female Orgasm; e.g.,
    http://philbio.typepad.com/philosophy_of_biology/2005/08/_orgasm_again_e.html )

    For starters, it’s important to keep in mind that lots of things are normative in some way or other without actually being normative for us. The rules of any game (no matter how silly, or uninteresting, or brutal), the laws of any state (no matter how inefficient, or unjust, or simply superfluous) — these possess some sort of normativity. But there is no inference from something have that sort of normativity to the conclusion that we ought conduct ourselves in accord with those norms. Indeed, any statement of the form “You should…” will have some sort of normativity; but only a small subset of such statements actually also have the kind of normativity that really places obligations on us.

    The best explanation for why it might appear that selective/persistence normativity is automatically normative for us, is that the discussion has focused too much on cases where the two sorts of normativity just happen to converge — i.e., cases like respiration or digestion, where well-functioning from the point of view of biology is indeed something that we also count as well-functioning from our point of view as persons. But once we step away from cases like those, we can begin to see that the persistence normativity qua persistence normativity fails to exert even the slightest demands on us as persons.

    Consider various hypotheses that have been offered in evolutionary psychology: that depression (e.g., Watson & Andrews), homicide (e.g., Buss), and even rape (e.g., Palmer & Thornhill) might be adaptations, and thereby qualify for the status of having persistence functions. It would be ludicrous to claim that therefore we ought to engage in such activities. It is entirely plausible that, when viewed from the standpoint of persistence functions, adultery will be normative, and homosexuality counter-normative. Yet, again, we would not want to say that therefore we should engage in the former, and should refrain from the latter. One more case: Lisa Lloyd’s book, mentioned earlier, argues that female orgasm has no evolutionary function; but denies that this means anything like women’s having orgams is normatively wrong or nugatory. In general, the people who make their living in offering explanations in terms of persistence functions have been at extreme pains to deny that there’s anything like an inference from persistence-function normativity to personal normativity.

    I should note that this is not an issue of whether or not anything can have intrinsic normativity. If you want to take these various systems, behaviors, programs, etc. to have intrinsic normativity of the biological sort, then go ahead — I don’t have a dog in that fight. And I’m happy with the idea that being a human person carries with it certain intrinsic normative obligations. I speak of “our goals” for cognition in order for my Craigian metaepistemology to be silent on the question of where those goals come from, but I am very open to the idea that, in fact, we are required to have certain of those goals, in virtue of being the sort of creatures we are. Indeed, I completely agree with every word of the end of your paper, so long as I don’t follow your use of “natural” to mean something like “can be cashed out in terms of explanatory functions”:

    “Knowledge and other positive epistemic statuses are worthy of pursuit by inquisitive creatures not (or not just) because they are instrumentally valuable. They may, of course, be instrumentally valuable – we need not disagree with Kornblith on that point. But that’s not the only thing that makes them worthy of pursuit for inquisitive creatures. What makes them worthy of pursuit for inquisitive creatures like ourselves is that, like health, friendship, and love, their attainment is partly constitutive of our well-being. Knowledge, and epistemic excellence more generally, is part of what constitutes the natural and valuable phenomenon of an inquisitive creature’s well-being.”

    What’s gone wrong is that you’ve tried to hypernaturalize our well-being. There’s just no need to explain this notion of well-being in anything like evolutionary terms. Yet by doing so, you open yourself to all sorts of troubles. First, on your view, epistemology takes on empirical debts that it can’t hope to repay (at least not for a long, long time), leaving open the kind of skeptical worries & odious possibilities I discussed earlier. To try to head off those sorts of worries, you have to concoct a framework in which our intuitions about funky cases can be citable & dispositive in a debate between empirical theories, even though you acknowledge that they can’t be evidence — leaving epistemologists in the odd position of being able to feel like they’re getting something done when they throw around their purely subjective estimates of prior probabilities, instead of doing what they ought to do, on your account, which is go learn the relevant science. And the idea that those subjective estimates are even vaguely relevant to the empirical questions can only be secured by your trying to make it out that we are ‘responsible’ for the cognitive systems in question; but that leads to the troubles I argued for above, for any notion of responsibility thin enough to allow us to count as responsible for cognition, is too thin to entail anything about persistence functions.

    So we can see all the worries that I have been raising for you here as coming ultimately from this one place: a confusion between natural selection’s normativity and the normativity of persons. Dissolve that confusion, and perhaps all the other false steps will no longer have even the illusory attractiveness that they seemed to.

  62. To put my point in the vocabulary that Weinberg has just used:

    Epistemic norms are “normative for us” (in Weinberg’s sense of that phrase) only if we are epistemic self-regulators, i.e., creatures who engage inquiry, or what Nishi Shah calls “doxastic deliberation”, and not just any old cognition.

    But epistemic norms are not “normative for us” (in Weinberg’s sense) if we are not epistemic self-regulators, e.g., babies or beasts. The normativity of epistemic norms for such critters (and for us for the earliest and sometimes also the latest parts of our lives) is no more “normy” than the normativity of circulation or respiration. So I’m not confusing these two kinds of normativity — I can state my point in terms of the distinction between them.

  63. My claim is that you’re confused between normative-as-in-persistence-function and normative-as-in-normative-for-persons. That is a different type of distinction than normative-on-nondeliberators vs. normative-on-deliberators. The former distinction is one of types of normativity; the latter is one of which beings the norms are or aren’t operative on. So one can have normative-as-in-persistence-function-on-deliberators (which I’ve been arguing is not of any real interest to us as cognitive agents); normative-as-in-persistence-function-on-nondeliberators (clearly not of any interest to us as cognitive agents); normative-as-in-normative-for-persons-on-nondeliberators (which Kornblith has much to say about); and finally normative-as-in-normative-for-persons-on-deliberators (which is what most epistemologists are interested in).

  64. I’m not confused between the two distinctions, but I am claiming that they’re aligned in the case of epistemic norms.

    There are persistence functions norms for lots of functional systems (circulation, respiration, brute cognition), and those are not “norms for us”. Once we become epistemic self-regulators (i.e., inquirers), we become responsible for our own cognition, and so it acquires a new persistence function, and the latter is normative for us, unlike the persistence function norms for circulation, etc.

  65. Ram, you’re going to have to explain to me how this comment from just a few posts ago:

    “What goes for circulation and respiration goes for cognition: the goodness is intrinsic, and we have a reason to want them, though this reason is independent of all of our ends and desires and goals” (emphasis added)

    can be consistent with

    “There are persistence functions norms for lots of functional systems (circulation, respiration, brute cognition), and those are not “norms for us”.”

    The more recent quote doesn’t look like it’s running these forms of normativity together, but that earlier one (which is more of a piece with the argumentation of the paper) sure does.

    Anyhow, what’s at stake here is how it might be that the persistence norms of cognition should have the status of being normative for us, given that persistence norms of pretty much everything else completely lack that status. The closest you’ve come is suggesting that we are responsible for their persistence, but as I’ve argued, there’s just no notion of “responsible” such that both (i) we are responsible for cognition, and (ii) responsibility for X entails anything about persistence functions of X. Minimal responsibility, as discussed earlier, will get us (i) without (ii), and more robust notions of control will give us (ii) without (i).

    I would go further, however, and argue that even under the (very counterfactual) circumstances in which we are in sufficient control of cognition that our own express goals might fix its persistence function, then it is still the goals and not the persistence function that is doing the real normative work. So even if we could find circumstances where the two types of normativity do converge, all that would mean is that persistence functions would be normatively-inert-but-at-least-pointing-in-the-right-direction instead of normatively-inert-and-so-we’d-better-ignore-them-because-they’re-pointing-in-the-wrong-direction.

  66. In the earlier post, I was using the phrase “reason to want them” in a sense that’s broader than what you call “norms for us”. The apparent discrepancy results from my attempt to switch to your terminology mid-thread.

    There’s no issue about persistence functions of brute cognition being “normative for us” (in your sense), since I claim that they’re not normative for us.

    Why is the persistence function of inquiry “normative for us”? Because we’re responsible for its persistence. To say that we’re responsible is not to imply that it’s intentional, or under our complete control. We’re responsible for lots of activities that aren’t intentional, or under our complete control. I don’t see what’s “minimal” about such responsibility.

  67. “In the earlier post, I was using the phrase “reason to want them” in a sense that’s broader than what you call “norms for us”” The confusion is at least as much my fault, then, since what I was trying to mean (and apparently not succeeding) by “normative for us” was something pretty much synonymous with, or at least entailing, what you were calling “providing us with a reason”. So my last few comments should be read as denying, in your terms, that persistence functions do anything at all like providing us with reasons. So I am flatly denying your assertion that the persistence function of circulation, for example, exerts no normative force on us whatsoever, and the several examples I discussed are meant to defend that denial; even if there is persistence function for some sort of rape program in our evolved psychology, that in no way provides us with even the thinnest sort of reason to exercise that program in accord with tha function.)

    “Why is the persistence function of inquiry “normative for us”? Because we’re responsible for its persistence.” As I argued above, there are many clear counterexamples to the thesis that such a thin notion of responsibility can generate anything like real normativity. In general, this notion makes us responsible for nigh-well every aspect of our social, economic, or political world, and indeed any aspect of our physical or biological world that is affected by our socio-economico-political world. That vast swath of responsibility already suggests that this notion is indeed pretty minimal, in that it places the bar for responsibility far below where any of our actual responsibility-attributing practices would place it; but put that concern aside. The deeper worry is that we would be answerable to the persistence functions of, well, the entire human world, even where those functions are normatively odious. That’s a direct upshot of the voting system, etc. cases.; and a threat of the cognition’s-persistence-function-might-be-brute-persuasion, etc. cases.

  68. Lions have a reason to gang up on the straggler in a herd of elephants, and suffocate it by biting its throat. This is a reason that has nothing to do with reasoning, or with rationality. But it’s the same sort of a “reason” that we have, to do whatever our “wiring diagram” calls for. I would say it’s an easily and massively overridable reason (as in the “rape program” case that you give above), but it’s so different from the kind of reasons that have to do with reasoning and rationality, that I don’t know if one can stand in the overriding relation to the other. That’s the only kind of epistemic “reason” that babies, beasts, and other epistemically non-responsible critters can have, on my view.

    I haven’t accepted your alleged counterexamples to the claim that responsibility (in the sense in which we’re responsible for the persistence of inquiry) suffices for normativity. If we participate in a social institution in a way that renders us even slightly responsible for its persistence, we are thereby binding ourselves by the norms of that institution — even if we aren’t conscious of it, even if aren’t happy with it, and even if we make an irresponsible effort to avoid responsibility by driving with a bumper sticker that says something like “don’t blame me for the institution, I voted against it”. This is the biggest cost of the world’s growing more interconnected — we have more blood on our hands than anyone of decent conscience can bear.

  69. Based on what you say about lions, babes, etc. it’s clear that you don’t take persistence functions simply qua persistence functions to provide anything like the sort of normativity that’s involved in epistemic evaluation. If we just agree on that — that there’s no connection between persistence-function normativity and epistemic normativity — then I won’t object if you want to continue to say that persistence functions provide something like a “reason”, as a sort of courtesy to them. (Though I do still suspect that it courts confusion to do so.)

    Similarly, you can attribute this slender sort of responsibility all you like (in a sort of progressive, awareness-of-our-interconnectedness-with-all-living-things kind of spirit, for which I have some sympathies). But it’s not going to do the work of transforming persistence function normativity into epistemic normativity. The relevance of the icky persistence functions here is not primarily as counterexamples to your notion as a notion of responsibility — which is why I said above to “put that concern aside”. What the cases serve to make clear, rather, is that this sort of responsibility just has nothing at all to do with actual epistemic normativity. It’s just preposterous to say that “reason in accord with modus ponens” (or any other candidate epistemic norm you like) has the same normative status as “vote in such a way as to serve your wealthy masters” and “enforce the linguistic rules so as to maintain sharp class boundaries”. That is, even if you establish some disappearingly thin, so-tenuous-as-to-maybe-not-even-count-as-overrideable sort of ‘normativity’ this way, it’s just not even a contender for being an account of actual epistemic normativity.

    And all that is in addition to the worries about metaepistemic skepticism & odious possibilities I argued for earlier. What your liberal conscience should want from your metaphilosophy is a way to say “these practices persist by serving a terrible function — we must change them, or perhaps even do away with them”. Yet your account would require us, rather, to say “we are responsible for these practices; therefore, we must serve their persistence functions, no matter what they may be.” I don’t think you want that conclusion, but so long as you insist that persistence functions play any role whatsoever as a source of or constraint upon the real normativity in our lives, it is not a conclusion you can avoid. Use a thicker notion of responsibility to ground epistemic normativity, and the icky cases might drop away, but so would cognition; try to understand epistemic normativity in terms of a thin notion of responsibility, and cognition is entirely of a piece with everything about our current social structures, including both the good and the unconscionable. I can’t see how that’s a result that you’d embrace, and I sure can’t see how it’s a result that makes this a very attractive way to naturalize epistemology.

    (There are also worries here that could be raised about distributing responsibility so widely that no individuals can ever be held accountable — if we’re all responsible for the disaster in NOLA, why should poor Mike Brown get fired for it? — and it might be interesting to think about what the corresponding epistemic versions of problems with distributed agency might look like.)

  70. “It’s just preposterous to say that “reason in accord with modus ponens” (or any other candidate epistemic norm you like) has the same normative status as “vote in such a way as to serve your wealthy masters” and “enforce the linguistic rules so as to maintain sharp class boundaries” People who vote in such a way as to serve their wealthy masters, and people who enforce linguistic rules in order to maintain sharp class boundaries, are, by and large (though with some notable exceptions), like lions who go for the straggler: they’re following the program, but they’re not deliberating (despite going through a pantomime of deliberation, in which they rehearse talking points that they’ve heard over and over.) Deliberation is not an altogether commonplace phenomenon even among adults, but belief-formation outside the context of deliberation is no more normative than respiration.

    “(There are also worries here that could be raised about distributing responsibility so widely that no individuals can ever be held accountable — if we’re all responsible for the disaster in NOLA, why should poor Mike Brown get fired for it? — and it might be interesting to think about what the corresponding epistemic versions of problems with distributed agency might look like.)” It doesn’t follow from our all being responsible that no one can reasonably be held to account. We’re not equally responsible.

  71. “…belief-formation outside the context of deliberation is no more normative than respiration.”

    But who says that the voter and the grammarian are unreflective? They might spend a fair of time thinking about who to vote for, or how to apply various of the grammar rules. Indeed, I expect that that’s pretty standard in both domains, and the other sorts of domains that generate relevant problem cases. If normativity is just persistence-function-plus-responsibility-plus-deliberation, then there’s no reason to think that they don’t have plenty of it. Maybe you have some very special form of deliberation in mind, such as deliberating over what you’d like the function of the practice to be, but then it will turn out that almost no one who isn’t an epistemologist will partake of cognitive normativity. (Also, do you want to deny that there are normative demands on our voting and writing, etc.?)

    So, at first (and still, in the paper), the source of normativity was the persistence function simpliciter, and deliberation, etc. came in as a way to try to say that the persistence functions for inquirers might be different than for brutes. But that didn’t work out. So then, earlier in this thread, it became persistence-functions-when-sprinkled-with-responsibility-dust. And that hasn’t worked out, either. And now you seem to be lobbying for normativity-as-persistence-functions-when-sprinkled-with-the-special-deliberative-brand-of-responsibility-dust-which-lets-you-ignore-the-actual-persistence-functions. I find it very plausible that deliberation brings its own kind of normativity in, but it sure looks like it’s something about deliberation, and more or less nothing to do with persistence functions, that’s doing all the normative work here. That is, in order for you to make this most recent move, you’ve basically got to yield the main point of contention here: that there’s no work to be done by persistence functions in an account of epistemic normativity.

  72. As I tried to make clear in the paper, the persistence function for inquiry differs from the persistence function of brute cognition. Inquiry just IS doxastic deliberation — I started using the latter term a few entries ago only because I thought it might signal a clearer distinction from brute cognition. And all deliberation, as I’ve been using the term, involves responsibility.

    I don’t see what it was about any of this that “didn’t work out”.

    In any act of doxastic deliberation, or course, there are some things (call it “the framework”, to coin a phrase) that are off the deliberative table. Our acceptance of those things is epistemically neither rational nor irrational — though it might be rational or irrational in other ways, and it might be epistemically good or bad depending upon whether or not it satsfies the persistence function of brute cognition.

    The normative demands on NONdeliberative behavior (be it voting, writing, or whatnot) are like the normative demands on lion behavior. The normative demands on deliberative behavior, however, are quite different. So this means that there are three cases: (1) someone deliberates well about whether to F, (2) someone deliberates badly about whether to F, or (3) someone doesn’t deliberate about whether to F (though they might, of course, pretend to deliberate about whether to F). Epistemic rational appraisal is appropriate to (1) and to (2), but not to (3) — the sort of appraisal appropriate to (3) is different.

    The norms of deliberation are like the norms of brute cognition in just one respect: they are both generated by persistence functions. Deliberation, unlike brute cognition, persists only because we make it persist — and we are responsible for doing so.

  73. Mike earlier (in comment no. 2) claimed as a counterexample to “E confirms H more than it does H’ iff. Pr(H/E) > Pr(H’/E)” these three propositions:

    1. Pr(H/E) = Pr(H), and

    2. Pr(H’/E)> Pr(H’), and

    3. Pr(H/E) > Pr(H’/E)

    However, it looks to me like 1 and 2 are incompatible, since

    P(H’|E) = 1 – P(H|E)
    = 1 – P(H) by 1
    = P(H’)

    so the counterexample fails. Perhaps somebody has already noticed this, but I’ve skimmed the discussion and can’t see this point made.

  74. In comment (2) above I said this,

    “It seems reasonably clear that it is not in general true that E confirms H more than it does H’ iff. Pr(H/E) >
    Pr(H’/E).” And that is rather obviously true. The degree of confirmation that E provides for H depends (at least) on the difference in value between the prior probability of H and the probability of H on E. The obvious fact is that
    Pr(H/E) might well exceed the value of Pr(H’/E) even if E provides no confirmation at all for H and considerable confirmation for H’. There are many cases in which this is true, but we need only one. Of course there arecases in which it is not true, but so what? My objection is to the principled attempt to move from Pr(H/E) > Pr(H’/E) to the conclusion that E provides more confirmation for H than for H’. That general claim is flatly false.

  75. Maybe a brief example. Let H = some tautology T and let H’= it will rain tomorrow. And finally let E = the weather guy predicted rain for tomorrow. It is pretty easy to see that,

    1. Pr(H/E) = Pr(H)= 1

    Given (1) E does not confirm H.

    2. Pr(H/E) > Pr(H’/E)

    The weather guy is good, but his predictions are less than certainly true.

    3. Pr(H’/E) > Pr(H’)

    The weather guy’s prediction confirms that it will rain
    tomorrow.

  76. Looks like I mistook which notational convention was in operation (I thought you were using the convention in which H’ is the negation of H). My mistake. Should have skimmed slower!

Leave a Reply

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