Are There Fundamental Epistemic Principles?

I’m interested in a question in the neighborhood of the title, but not exactly. Suppose we think of epistemic principles as (propositional analogues of) rules of belief formation and revision (including reasoning but not limited to it), and that there is a difference between belief change in accord with a rule and belief change that follows a rule. If this is the right way to think about epistemic principles, then we should expect that they have something to do with the distinction between propositional and doxastic justification: only when a rule is followed can that principle be used in an explanation of why a belief is justified (rather than simply having a content that is justified, while the state or event of believing is itself unjustified).

If we assume that rule-following is essential in this way to the distinction between propositional and doxastic justification, then I think we’ll have to give a negative answer to the intended question.

The rules we follow are typically local rather than universal: that look, when displayed by my wife, means anger; displayed by others, it may mean nothing. So too in cases of more complex reasoning. We use mental heuristics and rules of thumb rather than universal rules. And we don’t do this because we’ve arrived at the local rule by some combination of universal rule plus appreciation of the nature of our circumstances. Etiologically speaking, the local rules have priority.

What difference does this point make? It tells us something about what epistemologists are after when they attempt to articulate correct epistemic principles. They can’t be after principles that have imperatival analogues that play some actual explanatory role in belief formation and revision. Those principles are not ones that yield rules that a person needs to follow.

Call the position that disagrees, that holds that there are fundamental epistemic principles that have corresponding rules that we ought to follow the “Boghossian view” (the practice of forming an adjective from a name fails us here… sounds like the view of a follower of some unknown philosopher Boghoss–but that sounds like the result of a consonant-inversion game I used to play with my kids…).

There’s some connection here to van Fraassen’s attack on the Bayesian idea that we have a complete theory of evidence encoded in our conditional degrees of belief, a theory that tells us what adjustments to make (by conditionalization or Jeffreys conditionalization) given any possible future experience. But I’ll leave that connection and maybe come back to it at another time.

If the Boghossian view isn’t a correct account of the search for fundamental epistemic principles in epistemology, what is the alternative? One option is that there are rules that we must conform to in our cognitive practices, but it makes no difference whether we follow the rule or merely operate in accord with the rule. Another option is that epistemologists have been confused, trying both to construct a theory of justification using epistemic principles of various sorts while at the same time trying to construct principles that have corresponding rules that we must follow in order for our beliefs to be doxastically justified. Maybe there are other options as well?


Are There Fundamental Epistemic Principles? — 61 Comments

  1. The position you are pushing sounds like particularism in ethics. There are varieties of this position, but the general idea is that there are no true, informative moral rules. See Jonathan Dancy’s latest work for the fullest presentation.

  2. “. . .only when a rule is followed can that principle be used in an explanation of why a belief is justified. . .”

    I am justified in believing p on the basis of rule R only if I followed R in forming p (and not merely acted according to R). Is that close?
    It’s a little difficult to see why. Take inference rules as among the fundamental epistemic principles. If you are justified in believing (p & q) and claim that *so* you are justified in believing q, I’m sure I’d agree and explain this by appeal to simplification. Suppose you have no idea that there is such a rule and have no local substitute (who does?). It’s true (or so it seems) that the rule is nonetheless part of what justifies you in believing q.

  3. Hi Jon,

    It seems that the defender of the “Boghossian view” may try to push on the distinction between belief change in accord with a rule and belief change that follows a rule. I agree that people typically follow local rules, but it’s less clear that they don’t follow universal rules as well. It’s certainly the case that people aren’t aware of the universal rule; if so epistemology would be a lot easier! But mightn’t we follow rules that we aren’t aware of? We may follow a local rule that specifies belief given a certain look. But there’s a more abstract way of capturing that rule in virtue of the properties that such a “look” has (e.g., it seems to indicate the truth of the belief to you). We could then formulate a universal rule that abstracts away from the particulars of the local rule.

  4. Heath, maybe so, but I’m not convinced yet. The question here is about the role of rule-following, not one about whether there are true epistemic principles.

  5. Mike, yes, I would think &-E does have some explanatory role in case you cite, but the question is whether it does so in virtue of the individual following that rule.

  6. Ted, yes, rule-following is the central issue here. If you think that there is a host of rules that are being followed in a given case, that will help. But that sounds like a mistake, even if sometimes we are able to pull off multiple followings. On what grounds would we posit multiple followings? What would we need to posit such for, except to save the epistemological theory?

  7. Jon,

    There will be a generality problem here for rule-following. What is the actual rule that one followed? Is it, believe that my wife is mad given that look and this circumstance, or that look and its being Thursday… etc.? I think the oddity of following multiple rules arises when we consider rules that (roughly) are on the same level. (Please don’t ask about how to individuate levels…. 🙂 ) It doesn’t seem as odd to talk about following multiple rules when the rules are at different levels. If we jump up a level we can describe the rule as follows: believe that my wife is mad when it seems to me that she is. It seems plausible that I can follow both rules with a single case of belief-change. One is just more specific than another.
    About the need to posit such multiple followings, it seems to be needed to explain how the same person can follow the same rule on a different occasion or how two or more persons can follow the same rule. There may be certain higher-order properties that guide belief formation, but these properties can supervene on different lower-level properties. The lower-level properties are the ones that figure in the local rules; the higher-level properties are the ones that figure in the universal rules.

  8. “…the question is whether it does so in virtue of the individual following that rule.”

    Yes, right. But the hypothesis was that the person in question is not following the universal rule (simplification) and is not following a local rule (whatever that might be). He is acting according to the universal rule. I think this is just what happens when people reason in an intuitive way. There just aren’t aware they’re following any rules at all. It is typically news in a logic class that there are such rules. Still the rules play a role in justification.

  9. Mike, if there is a rule being followed in the case of simple deductive inferences, it won’t be a general one having to do with the logical behavior of the connective. I guess we can always get one: when faced with these premises in precisely these circumstances, believe this. Not exactly what the Boghossians will need, I would say.

  10. Ted, I think the reference to the generality problem is a red herring. The question of which rule one is following ought to be an empirical one, just as the question of what the laws of nature are is an empirical one.

    So consider the question of why a particular event occurred, and the kind of DN-explanation one might favor. You know the antecedent circumstances, and in the place of the laws, you have a number of different generalizations all of which, together with the antecedent circumstances, imply the event in question. The DN-theorist says: the real explanation is the one that contains the real law. Your position would seem to be: the real explanation is the one that contains all the generalizations, generating a logically overdetermined conclusion.

    Another example: Kripke wants to know, when you put 27 and 51 together to get 78, whether you are adding or quadding. You say, “both!” “No, no, no, you don’t understand the issue then.” I expect defenders of rule following in epistemology or ethics would say the same. Kant judges the moral worth of your actions depending on the maxim involved–sounds like you want to throw in as many maxims as you can and claim that you are acting on the basis of all of them.

    So take your last example: following the rule of believing that your wife is mad when she seems to you to be. What’s the difference between merely coming to the belief that your wife is mad when it seems to you that she is, and following the rule in question on a particular occasion?

  11. Jon,

    I agree with your discussion of the Kripke example and the Kant example. The “Boghossian” position I put up for discussion was intended to be yield that verdict. In particular, when considering rules at the same level it doesn’t make sense to be guided by multiple rules, as evidenced by the Kripke point. However, if we can distinguish specific rules from more general rules, it doesn’t seem incoherent to suppose one can follow a specific rule and also follow a more general rule at the same time. Take a Kantian example. I am guided by the maxim to help orphans. I also am guided by the rule to help the unfortunate. Can’t I follow both rules when helping an orphan? The one is just more specific than the other.

    About your last question, I am not sure what you’re driving at. I’d explain the difference between “my merely coming to the belief that your wife is mad when it seems to you that she is, and following the rule in question on a particular occasion” in terms of appreciating the evidence. In the former case I don’t appreciate the evidence and come to the belief by happenstance. In the latter case, I appreciate the evidence. Following an evidential-rule seems to require being guided by the evidence.

  12. Jon,

    I don’t think I’m anxious to defend the Boghossians. In any case, I am not now arguing that such a view is right. I am worried rather about another claim you make. The claim is this (I’ll call it C):
    C. “. . .only when a rule is followed can that principle
    be used in an explanation of why a belief is
    justified. . .”
    What you have in mind by C (I’m suggesting, not reading minds) is C*,
    C*. I am justified in believing p on the basis of rule R
    only if I followed R in forming p (and not merely
    acted according to R in forming p).
    I offered an example in which someone S was *not* following a rule R (where R = simplification) and yet simplification justified his inference. From this I wanted to conclude not that the Boghossian’s are right, but that you can be justified in believing a proposition (in this case q) on the basis (in part) of a rule that you are not following.

  13. Ted, the Kripke example was about same level cases, but the DN example was about multi-level cases. I won’t return to it though, but here’s the point. In the usual case, following a rule will affect the causal/explanatory story we tell about a case. If you’ve got multiple rule-followings, that would mean that each of them contributes a little to the causal/explanatory story (or that there is real overdetermination here). Either option is unsavory for the usual case.

    So, let’s suppose that anytime I believe my wife is angry when it seems that she is, I’m following the rule that says, “when she seems angry, believe it!” I’m also, then, following the rule, “when she seems angry and 2+2=4, believe she’s angry.” I am, most assuredly, not following that rule.

  14. Ah, Mike, I see now. My gloss that you’re taking issue with was meant to characterize what I take the Boghossian position to want to, or have to, endorse. I too doubt it’s truth, though I hesitate because I don’t really have a firm understanding of what it is to follow a rule. I think there are clear cases where I am following a rule, but when I look out at the oak tree in my back yard and believe that it’s an oak tree, I don’t really know if I’m following any rule at all. I tend to think I’m not, but maybe there’s some specialized sense available in which I am.

  15. Jon,

    One last thing. My initial reaction to the DN-example was that it wasn’t a problem. The view, I think, would be that there are multiple genuine explanations for an event’s occurrence ordered by the level of specificity of the laws and conditions. This doesn’t seem to imply over-determination. We don’t have two independent processes leading to an event’s occurrence. We have one process that can be described at different levels.

    I agree with your example; I would have thought that a same-levels case. The kind of case I was thinking about is where the rules are “When it seems to me that my wife is mad, believe she is” and “When she has that look, believe she’s angry.” If we have two rules here, it seems I can follow both in one case. If only I knew what rule-following was…. In any case, I have to stop because my wife is giving me that look!

  16. Ah, Ted, you seem committed to the view that there’s a solution out there to the problem of mental causation–if the low-level physical explanation is adequate, what work is there left for the mental? I think the literature here should lead you to be pessimistic…

    Notice that you don’t claim to be happy with following the rule that if your wife gives you that look and 2+2=4, then believe that she’s angry. That’s good, you shouldn’t be. But, as you can tell, there’s a recipe here for constructing an infinite number of rules.

    There also a further problem, but I’m going to make my next post about it…

  17. There seem to be three issues here, clearly separable from each other. The first is: are there epistemic rules? Answer No, and you are probably a particularist.
    The second is: how general should the rules be? Answer: very general, and you are a full generalist. Answer: general with exception, and ceteris paribus clauses, and you are a hedged generalist (or moderate particularist – Mark Lance and Maggie Little in ethics are good examples)
    The third: should the thinker be guided by them, or is it enough that her practice merely de facto accords with them?
    My own preference yes, hedged generalism. (We do go by rules quite often. Take the anger example: there are some facial expressions that are almost unmistakable signs of anger. Others are unmistakable in normal circumstances).
    A naive, unreflective thinker is in clear if her practice accords with rules.
    Full reflective justification involve awareness of the relevant rules.

  18. I think that Ted must be right about the possibility of rules operating at different levels (whether or not the more persnickety metaphysicians have worked out exactly how to capture that fact). One test for when this is occurring is when it is the case that one rule can be said to be complied with in virtue of the other’s being complied with. E.g., at the relevant time of the year, I am guided by both the rules “obey the laws of the United States” and “file your taxes before 4/15” — but I am being guided by the former in part in virtue of being guided by the latter.

    As I started to type, I was thinking that this possibility of multi-level rule-following has much bearing on the question that initiated the original post. But now I’m not so sure. What if we are guided by some general rules like, say, “believe in accordance with your evidence” — but we cannot follow those rules unless we also have some fairly particular rules about cases of good evidence and the like? Even if we cannot follow the higher-level rules directly (i.e., without having some lower-level particularist rules to follow as well), does that mean that those rules are nonetheless not fundamental?

    I guess it depends a lot on what the relevant notion of fundamentality is, which is something I also wanted to post a comment on to ask Jon. If my my hypothesis were correct, it seems to me that the general principles might be normatively fundamental (i.e., not depending on some other principles for their normative force), but in some sense metaphysically non-fundamental (i.e., can only be complied with in virtue of complying with still further rules). I guess they would also be epistemically fairly non-fundamental (i.e., we could only know them indirectly, since as has been noted already, if they were immediately available to introspection, epistemology would be a lot easier!)

  19. It seems to me that the view that Jon is articulating (though perhaps not advocating?) is essentially right — belief change isn’t “governed by rules,” because rules are explanatorily impotent.

    Perhaps this is a useful analogy: consider plants that turn to follow the sun during the course of the day. One could try to explain this in terms of a rule according to which the plants are permitted or obligated turn in the direction of illmination, but that would be a strange explanation. The normative terminology seems to be irrelevant; what is doing the explanatory work is simply the plants’ genetic disposition to follow the sunlight.

    Similarly, what explains belief change is not rules of inference, but simply inferential dispositions.

  20. Jonathan:

    There are very interesting worries about when it is true,

    “that one rule can be said to be complied with in virtue of the other’s being complied with.. .”

    Two examples. (i) Neo-utilitarian moral theories (F.Feldman’s, for instance) would have agents bring about the best possible history from time t (now) onward. Suppose A at t is part of the best history h. It is often true that the same action A is part of the worst history h’ (it depends largely on what you do after A at t). But, though A is part of the worst history, performing A at t needn’t entail that you’re following the rule “realize the worst history”. It seems perfectly reasonable to perform A and deny that you are following the rule “realize the worst history”.
    (ii) You might follow the rule “you ought to do A” or OA, and not follow the rule O(A & B) since there might be no rule O(A&B). For instance in cases where O(A&B)-> (A&B) and its true that ~(A&B).

  21. Mike: That’s of course right — and illustrates the importance of the distinction between rule-accordance and rule-following. Since one can be acting in accord with a rule without following that rule, one can (to use your example) act in accord with the rule “actualize the worst possible history” without following that rule. So, perhaps the problem was my use of “comply”? I meant for “comply with a rule” to be synonymous with “follow the rule”, and thus not be synonymous with “be in accord with the rule”, but maybe “comply” is ambiguous (or even tilts the wrong way!) So, let’s pretend I used “follow the rule” consistently throughout my earlier post… if one did indeed find/replace all my “comply with”s to “follow”s, do you think there’s still that kind of problem?

    Stephen: You seem to be committed to the claim that all rules are explantorily impotent, but that doesn’t seem to be true. It is true of many creatures, like plants, but it’s not true of humans. Indeed, I think my earlier example of being guided by rules about tax law is an example where the rules themselves are quite explanatorily relevant to my behaviors. So there’s at least the possibility, then, of our inferential dispositions being actually governed by the rules themselves, the same as my tax-filing dispositions are. I take it to be part of Jon’s point, then, that as a matter of contingent fact, we are not so constituted cognitively as to actualize that particular possibility.

  22. Jonathan, as I understand the Boghossian view, we want (i) epistemic principles that are exceptionless and have whatever status in terms of necessity and a prioricity such principles ought to have (!) and (ii) the related rules are ones that we follow in belief formation and revision. If we get those things, we’ve got fundamental enough principles–nonfundamental means, I think, contextual or relative rules. So, whatever rules we follow, they ought to have explanatory force, not just evaluative force. That leaves open the possibility of multiple rules being followed, but it also leaves open, I would think, that only one of your taxation rules is being followed while the other I’m only acting in accord with. So, I think you’re right that multiple followings are possible, but only when the multiplicity in question is explanatorily relevant.

    I’m going to post again on this subject to try to make clearer what is very murky in this post: what’s the relationship between epistemic principles and rules. I’m thinking that getting clear on this will sort out the mess I perceive in talking about distinctions between subjective and objective justification, justification versus blamelessness, and talk about primary and secondary propriety with respect to some rule or other. We’ll see…

  23. Jonathan, you say this,
    “One test for when this is occurring is when it is the case that one rule can be said to be complied with in virtue of the other’s being complied with. . .”
    I guess I don’t know what you mean by “in virtue of”. I thought you meant something like this:
    If I am following rule R that demands A and A is also demanded by R’, then I am following R’ “in virtue of” following R’.
    But you are not saying that, it seems. Since I might follow the requirement to do A in “realize the best history” and in virtue of doing so NOT follow the rule “realize the worst history”, though the latter also requires A. Maybe what you mean is that I can follow rule R (“Stay to your right”) and follow rule R’ (“Don’t stay to your left”) at the same time. But that seems close to trivial, so I’m guessing that’s not it either. In any case, here I am not following one in virtue of following the other: I’m just following both.

  24. I guess it’s pretty obvious, but this,
    “If I am following rule R that demands A and A is also demanded by R’, then I am following R’ “in virtue of” following R'”,
    should have been this,
    “If I am following rule R that demands A and A is also demanded by R’, then I am following R’ “in virtue of” following R.

  25. Is a rule not a definer of conditions for when to do *this* and when to do *that*–or, in epistemology, for when to believe? Since the rules “when my wife has that look, believe she’s mad” and “when my wife has that look and 2+2=4, believe she’s mad” dictate exactly the same conditions under which you’re to believe your wife is mad, and since you clearly take the two to be different rules, one of which you’re following and the other of which you’re not, I have to wonder: Must you be aware of the semantic content of a rule in order to be said to follow it? Must you be aware of its sense and not merely its denotation in order to follow it?

    Keith Brian Johnson

  26. Keith, no you don’t have to be aware of a rule to be following it. If you mean the Fregean distinction between sense and reference, you don’t need to be aware of either its sense or its denotation to follow it.

  27. The notion I had in mind — and which I figured was at least something like what Ted was saying earlier — was a distinction between higher-order and lower-order rules. Many, perhaps most, rules are merely first-order, telling us what to do under various circumstances. But some rules — and, I suspect, at least a few very important ones — are second-order, telling us not what to do per se, but rather what rules we should look to follow in various circumstances.

    So the rule, “pay your taxes on 4/15” is a first-order rule. But part of the reason I follow that rule (or, um, at least try to do so) is that I am trying more generally to follow the rule “obey the laws of the United States”, and I believe that the tax rule is one of those laws. But that “obey the law” rule is higher-order — it doesn’t tell me what those laws are, but only to find out what they are, and to govern myself by them. And thus I can only follow that rule by following some other rules, namely, the rules that are the laws of the land.

    So to explain my behavior, you have to appeal not just to the first-order rule, but also to the second-order rule (and, I suppose, my belief about what the laws are here).

    And that’s the sense in which I can follow one rule in virtue of following another: since I can only follow the higher-order rule by following some other rules, it is in virtue of following the other rules that I follow the higher-order rule.

  28. My point is that if the two rules specify the same conditions for belief, and if you aren’t aware of their semantic difference, then it seems that if you’re following one rule then you’re following both rules (or possibly you’re not following any rule, but are simply acting [believing] in accord with both). But you deny that. If, instead, you are aware of their semantic difference, then presumably you know which of the two you’re following, and only that one plays a role in the causal/explanatory story; you’re merely complying with, or acting (believing) in accord with, the other one. But you deny such awareness. I gather there’s an option I’m missing?

    Keith Brian Johnson

  29. What rule is being followed depends on the kind of cognitive architecture in place. Just think of it like any ordinary functional entity: it takes input and generates output. The heart, the kidneys, the lungs all follow rules in this sense. The rule the lungs function according to is something like: intake air with oxygen, output carbon dioxide. The governing principle isn’t: intake air with oxygen and, if 2+2=4, output carbon dioxide. Saying exactly why the latter principle isn’t the one doing the work is a complicated matter, but that doesn’t affect the point in question.

    Same with belief formation and revision: the mind functions in a certain way, and not every rule to which one conforms is a rule that plays an explanatory role in describing the system in question.

  30. Jon, I’d be surprised if this longer rule (for lungs) plays no explanatory role in the example you describe,

    “intake air with oxygen and, if 2+2=4, output carbon dioxide”

    The only worlds at which 2+2 is not 4 are impossible worlds. But the rule,

    “intake air with oxygen and DO NOT output carbon dioxide”

    is true for lungs at impossible worlds (since everything is true there). And that is a rule (I assume) you believe does not hold for lungs. So the abbreviated rule,

    “intake air with oxygen and output carbon dioxide”

    does seem to assume that we’re talking about 2+2=4-worlds. The longer rule just makes this explicit.

  31. Which rules have true antecedents isn’t the issue from the explanatory point of view. From that point of view, it is what is doing causal/explanatory work in the system in question.

  32. It seems to me that a “intake air with oxygen, output carbon dioxide,” is not a rule, but rather an instruction. A rule has full blown normative content stating what something or someone is permitted or obligated to do under certain circumstances. It strikes me as odd to explain the activity of the lungs by saying that, having taken in oxygen, they are permitted or obligated to output carbon dioxide. The lungs aren’t “rule governed” in this sense.

    But it seems equally odd to say that the activity of the lungs is explained by the instruction “intake air with oxygen, output carbon dioxide.” Rather, the lungs have a wired-in disposition to behave in a way that can be described by that instruction. It is the wired-in disposition, not the instruction, that does the explanatory work (or at any rate identifying this disposition is the doorway to an explanation that ultimately gets spelled out at a deeper physiological level).

    While there might be instances of belief change that actually can be explained by reference to a rule specifiying what one is permitted or obligated to believe in certain circumstances, these presumably involve unique and specialized circumstances (such as Jonathan’s tax law example). In most instances I suspect it is more productive to bypass such rules and go straight to dispositions as the most salient explanatory factor.

    And it does seem likely that within a disposition-based framework of this sort, general epistemolgoical principles would play a very minor role.

  33. Stephen, first, the rules imagined have imperatival consequents; only the corresponding principle has the concept of obligation in the consequent. So we can discuss the idea of rule-following without imagining any corresponding obligation.

    I think as well that the thought that some troublesome metaphysical implications of rule-talk can be avoided by disposition-talk is not going to pan out. It’s not as if dispositions are somehow scientifically respectable in ways that instructions or rules are not. All this talk, if things go well, will need to be explicable by more basic properties; so not much is gained by substituting dispositions for rule-following. In fact, fans of rule-following might be perfectly happy identifying rule-following with dispositions of a certain sort.

  34. That probably is right. But I guess I’m trying to make two points: 1) If one invokes rules to explain rule-following, one has to invoke dispositions as well, but not vice versa. 2) If one avoids rules and explains cognition in terms of dispositions, intuitively one would expect the relevant dispositions in most instances to be highly particularized; general epistemic principles would be only peripherally involved in cognition, if at all.

    On your sencond paragraph, I wonder if it always is necessary to explain dispositional properties in terms of more basic underlying categorical properties, as you suggest. It seems to me that in at least some instances “bare dispositions” can be explanatorily informative by themselves. Consider fragility. If a wine glass shatters because one drops a brick on it, one doesn’t invoke fragility to explain that. But if the wine glass breaks after it tips over, then one does invoke fragility. So fragility seems to be capable of doing at least some modest explanatory work without first having to reduce it to underlying nondispositional properties (though presumably such a reduction would in principle be possible here).

    There’s a fascinating argument in Thomas Holden’s recent book “The Architecture of Matter” to the effect that dispositional properties can’t in general be explained in terms of underlying categorical properties. One would need to invoke a second dispositional property to explain how the postulated categorical property gives rise to the disposition in question; this second dispositional property would in turn require a further underlying categorical property, and so on.

  35. I’m very tempted to the view that the hope of explaining dispositions in terms of categorical properties is a lost cause, too. I’m not opposed to explanations by dispositions, by the way; I was only meaning to echo explanatory orthodoxy about what fundamental explanations have to look like. And I also don’t intend to commit to views about the relationships between codes/rules/instructions/dispositions, etc.; rather, I’m just trying to give the Boghossians as much as I can in order to see whether the closing game they need can be carried out.

  36. Here’s a question: if one is to be a Boghossian, must one be a representational realist? Clearly the easiest way to draw the following a rule/merely in accord with the rule distinction, is to cash out rule-following in terms of the agents having actual representations of the rules in their heads, which then play a causal role in their actions. Though I am friendly to this particular realism, it’s certainly not the case that everyone is (and I think varieties of antirepresentationalism are growing in philosophy of cognitive science). So is there a good way to make the distinction, if you’re not going to do so in terms of causally efficacious representations of rules?

  37. You noticed, Jonathan, that I’ve been fudging on this point! I’ve tried not to saddle the view with representationalism, since I would have thought that the causal/explanatory relations might go without them, as in the case of ordinary biological functions (though one then has to say something more about what makes something a rule…). It is, as you note, much easier to describe if we just adopt representational language, but I didn’t want to force the view to do so. But maybe there’s an argument somewhere that the view is committed to this…

  38. Wrt representational realism: It seems to me that there are two possible questions to ask that need to be separated. (1) Is an epistemic agent doxastically justified in her belief that p if she represents to herself (correct) rules of belief formation, and follows them? Versus (2) Is an epistemic agent doxastically justified in her belief that p if she follows those rules, in some other, non-representational, perhaps unconscious, fashion? One might give very different answers to these questions; it involves what doxastic justification requires. E.g. a proper functionalist might answer both questions in the affirmative, but a more internalist approach might say yes to the first but no to the second.

  39. Heath, there may be some versions of internalism that draw the line where you draw it, but they are pretty unimpressive versions of the view anyway. The internalist stance may be more wedded to the idea that rule-following is central to doxastic justification, though I’m stll not convinced (since I want to know more about rules and about following…).

  40. Heath, I basically agree with your point, but we need to avoid conflating “represented explicitly in conscious cognition” with “represented in cognition”. The former entails the latter, but not vice-versa, since the rules might be really represented, but unconsciously. So I’d rather parse out the two questions as, “Is the agent justified on the basis of following a represented rule?” and “If the answer to the first question is yes, what further conditions must apply to that representation?” Perhaps none; perhaps being tokened consciously; perhaps (as I think Boghossian has in mind) unconsciously-but-meaning-constitutively.

    Jon, I thought I had a long objection to appealing to proper function to make the relevant distinction, but on reflection I decided it didn’t work. So I’ll have to think on it some more. But I will note this: in systems that do have causally-efficacious representations, it is possible for the represented principles to differ from the proper-function principles. E.g., suppose that a device is required to calculate x-squared, but for some strange reason (a drug-addled engineer, maybe) it does so by iteratively adding x/2 + 1, x/2 + 2 … + (x/2 + x). It then doubles that sum; and then subtracts x; and finally divides by 2. [I _think_ that works….] Its proper function is f(x) = x2, but its representational function is, um, a bit more complicated. That the representational and proper-functional accounts of principles can diverge gives some reason to think we need to figure out why the Boghossian might need to pick one or the other.

  41. Jon, I’m pretty sure I don’t follow this,
    “Which rules have true antecedents isn’t the issue from the explanatory point of view. From that point of view, it is what is doing causal/explanatory work in the system in question”
    I was suggesting, in effect, that certain counterexample to the claim X is following rule R seem to be implicitly precluded. So for instance, someone might offer the counterexample to the claim that one’s lungs are following the rule “intake air with oxygen and output carbon dioxide” that in worlds where 2+2=5, lungs follow the rule “intake air with oxygen and do not output carbon dioxide”. This is true but I thought you’d want to say those worlds are irrelevant to the question of which rule the lungs are following. Other worlds are irrelevant too, such as worlds in which causal laws are different, and so on. How do you rule these worlds out? One way is to make explicit that the rule is relativized to certain worlds by building into those rules the relevant restrictions.

  42. Mike, I was claiming that the addition of 2+2=4 to the antecedent of a rule will only introduce a factor that plays no explanatory role in the generating the output in question. So if we compare “In C, output O” with “When in C while 2+2=4, output O” the second rule would be relevant only if the math claim played a role in the causal/explanatory story about the production of O. Since, I think, it won’t, one can follow the first rule without following the second.

  43. Note that the particular ways that funny rules can get strained off will depend a lot on what account of rule-following we’re working with. The representationalist about rule-following can do so very easily, since the representations are individuated in part by their syntax: “if p then q” and “if (p & 2+2=4) then q)” are thus different rules.

    The proper functionalist will have a slightly harder time. I am sympathetic to Jon’s appeal to explanatory relevance, but I’m not 100% sure that it’s best to do so in terms of causation — can it handle rules with negated antecedents? (E.g., “if there are no predators in the area, then make your mating call.”) I’m not opposed to granting causal efficacy to negative facts, but I worry that doing so here would vitiate the move against funny rules. That is, the broader the class of potentially causally relevant facts, the narrower the class of rules eliminated by the appeal to causation.

    Consider the alternate rule, “In C, if the lungs aren’t completely full of water, then output O”. I don’t think a causal principle will help eliminate this rival. But there are other explanation-based resources available. In these cases, we can see the conditionals as operating with a particular kind of modal force, And the possibility-space that underlies that modal force simply leaves out all sorts of unusual (even if nomologically possible, or even actually instantiated!) environments. So 2+2=5-environments, and lungs-filled-with-water-environments, are thus not under consideration. And so it is explanatorily otiose to include them in the statement of the rules themselves.

  44. Jon, I don’t see it. Here’s a different example. I ask what rule you’re following and you say it’s rule R.
    R. Do not lend money to other players.
    And I say, you mean you’re following R*, don’t you?
    R*. When playing Monopoly, do not lend money to other players.
    You say, ‘but adding that antecedent adds nothing to the explanation of what I am doing’. But I can’t see why not.
    R* is more narrow than R (R* is restricted to Monopoly-worlds) and it is R* that explains what you’re doing; you’re following a Monopoly rule, not a chess rule. In the same way the more narrow rule for lungs is what explains what the lungs are “doing” (i.e., the rule that is restricted to 2+2=4-worlds).

  45. Jonathan, nice examples. I think you’re right that explanatory notions will be the key ones, and I slipped in talk of causal notions just because… I don’t know, maybe because I feel lost at sea if causation and explanation come apart (very much?…). In any case, you’re right that causation alone won’t rule out your alternative lungs rule–the causal route is the same for both rules. If I were attracted to the usual ways of setting aside the Gettier problem, I’d say, “yeah, explanation is a mess, as much a mess as the Gettier problem; so let’s just pretend we’ve got that taken care of in some way and move on.” Terrible, terrible.

  46. Mike, maybe your worry is much like Jonathan’s? In any given case, you might be following R and you might be following R*–it just depends on the what the correct explanation of your behavior is. The causal path from input to output will be the same in the two cases, I would think, but one explanation is preferable to the other–just don’t ask me to say how! That problem, I take it, is the same worry as Jonathan’s, and without a solution to the multitude of problems in the theory of explanation, I have no defense of my view. By the way, adding the antecedent you add can do explanatory work–it would indicate the presence of dispositions that the other rule doesn’t. The presence of such dispositions would be explanatorily relevant in a way that, say, the difference between mathematically possible and mathematically impossible claims would not be.

  47. I’ve not expressed this well, I suppose. I’m not jumping on Jonathan’s bandwagon. I take it that Jonathan means to convey that ‘2+2=4’ is, as an addition, “explanatorily otiose” in some metaphysical sense. I deny it. It is likely to be explanatorily otiose, epistemically. But I don’t find that especially interesting or relevant. The epistemic version amounts to saying, “that’s the rule strictly, but everybody knows that”. I’m not saying anything about what everyone knows. I don’t know what everyone knows. My claim is that the addition is metaphysically relevant even if epistemically obvious. I’m pretty sure this is not just repeating what’s been noted.

  48. Mike, I can sort of see what you’re thinking, but not quite. I see that you want the math truth to be part of the explanation in some metaphysical sense, and explain away the denial that it is part of the explanation by appeal to some distinction between metaphysical and epistemic relevance. I don’t know what that distinction would be, however. As I think of explanations, they are not subject-dependent but are rather as objective as causation itself is. In that sense, I don’t see why trivial facts of arithmetic would be explanatorily relevant to what beliefs people form. I’m not saying they *couldn’t* be relevant, but I don’t see why we should think that they are always relevant.

  49. It seems to me that the 2+2=4 example just illustrates that whatever rules we follow, we follow them under the assumption that “the world is this way.” While he’s married to his wife, while gravity continues to operate, while houses don’t spontaneously crumble–as long as the world is the way it is, has been, and is expected to be–he will follow the rule, “If my wife looks at me this way, believe she’s angry.” It’s all backdrop; and in Jon’s terms, what’s merely backdrop does no explanatory work, except to the extent that it says, “Go ahead. Do what’s normal.” But the Monopoly example looks both similar and different. It’s stating assumed backdrop, yes; but in this case it’s not just part of specifying “the way the world is,” but is rather specifying special conditions under which the rule applies. The statement of special conditions surely has explanatory power.

    I was wrong about awareness, but intentionality has to play a role in explaining rule-following, it seems to me. Imagine Bugs Bunny walking along the road. Ahead, the road forks: left, to Daffy Duck’s house; right, to Yosemite Sam’s house. Before Bugs reaches the fork, which path is he following? Well, ask him! He might be aware of going to Daffy Duck’s, but maybe he simply never visits Yosemite Sam and is following the road to Daffy Duck’s out of habit. But if you ask him, he’ll be able to tell you which road he’s following. That the “two” roads to Daffy’s and to Sam’s overlap for a while doesn’t matter, unless Bugs, when you ask him, can’t tell you whom he intends to visit. Maybe he’s just out on a walk. But then is he really following either path? Or do we withhold judgment until he makes up his mind?

    I don’t know how happy I am with dispositions, but something intentional has to be invoked, it seems to me.

    Keith Brian Johnson

  50. Jon’s right that it’s not a matter of obviousness, but of what is actually, as a matter of fact, playing a role in the explanation. As Keith points out, explanations only make sense against a backdrop of a bunch of stuff that’s held fixed. And you’re not going to be able to list off all of that stuff, or even come close to doing so. After all, if it makes sense to bring in 2+2=4, then it’ll make sense to bring in all the truths of arithmetic, and probably logic as well, and indeed all the sorts of stuff I was gesturing at in my previous comment. And this list-resistant class of stuff-held-fixed is not the same as the class of the obvious. There are obvious things that are not among the held-fixed: that I exist, for example. And in many cases there are many facts among the held-fixed that are not obvious: scientific explanation takes a lot of hard-won scientific knowledge as fixed, none of which would count as obvious. (All contemporary biology operates with a whole lot of the theory of natural selection held fixed, for example.)

    So, 2+2=4 (inter multa alia) is (metaphysically) otiose in most explanations because it is among the held-fixed.

    Keith, I don’t see why it has to be something intentional that explains the path, and indeed the case as you describe seems to me to be a counterexample to any such intentionality requirement — as you say, maybe he’s taking one path instead of the other out of habit! In such cases, I don’t see why anything intentional need be operative. Sure, sometimes he can tell you what he’s doing if you ask him; but not every time. My own phenomenology of such cases is discovering that, say, I’ve accidentally driven on a path different than the one I originally meant to be on, and then I have to reason from the outside, “Hm, this is the way to get to the old school — I must have started heading there on autopilot.” (Maybe I’m taking habit to be non-intentional, but you take it to be intentional?)

  51. “And you’re not going to be able to list off all of that stuff, or even come close to doing so. After all, if it makes sense to bring in 2+2=4, then it’ll make sense to bring in all the truths of arithmetic, and probably logic as well . . .”

    This looks like an entirely epistemological point to me. Effectively, it would take omniscience to produce the list. Ok, yes. But that doesn’t tell us that the listed information is metaphysically irrelevant to which rule I’m following. The arithmetical qualification on the rule, I concede, is the limiting case. It assumes that there are worlds in which 2+2 = 5 (i.e. impossible worlds). But if you’ll grant for a moment that there are impossible worlds, then even the arithmentical qualification is relevant to what rule (say) the lungs are following. The rule is not perfectly general, but rather one that applies in a restricted set of circumstances. All of this is consistent with saying that it is a practical impossiblity to specify everything that is relevant to which rule is being followed. Yes, right. It is consistent with saying that there are facts that are “held-fixed”. I agree. But practical and epistemological considerations aside, this information qualifies the rule in ways that we’d be disposed to acknowledge when pressed to answer which rule, strictly, we’re following. The fact is that we’re never pressed that hard about it.

  52. It’s not a problem of omniscience; it’s a problem that you’ll need to countenance conditional sentences with infinitely-long antecedents, and the problems that presents are neither merely practical ones nor (to steal from Russell) a matter of biological limitation.

    I don’t want to argue that there can’t be some sort of useful transcendental notion of “explanation” that uses only philosophical machinery, so maybe you’re right about some legitimate notion of “explanation” that works that way. But it’s just not a notion that is extant in science, and in particular not a notion that is used in determining proper functions in biology.

    Note that I’ve agreed with you that the rule should be interpreted as applying only in limited circumstances — I just don’t think that you should build those circumstances into the rule itself. Rather, if we think of the rule as being of the form “[](p->q)”, I am advocating building the limitations into the ‘normal environments’ semantics of the modal operator. One way to see why it should be done this way is that two theorists can agree entirely as to all the circumstances under which p will or will not produce q in the system, but disagree as to whether some particular circumstance r is one of the normal environments (though agreeing that when r, p will not produce q). They would thus be disagreeing about the status of the rule “if p then q” for the system in question: the theorist who thinks r is among the normal environments will argue that r situations are counterexamples to the rule, whereas the other theorist will think r situations irrelevant to the rule’s status. I don’t see how the ‘pack it all into the antecedent’ account can make sense of this dispute — in terms of what’s ‘strictly’ going on, they are in complete agreement, and they do not take their disagreement to be one of merely practical matters. And I’d note that disputes like these are not uncommon in fields like biology and psychology, in which the developmental data (e.g.) can be agreed upon, but what then gets argued is the explanatory relevance of the fact that the organism fails to develop a given trait in some funky experimental environment.

  53. “It’s not a problem of omniscience; it’s a problem that you’ll need to countenance conditional sentences with infinitely-long antecedents..” I guess I should be more worried about this than I am. My position is not that we should hurry-up and pack all this information into as many antecendents as we can find. My position is that we *can* do this, if we are pressed about specifying the rule more precisely. And there are certainly logics in which infinite conjunctions are perfectly sensible. So, that doesn’t seem like an insurmountable problem. Still approaching the specification of rules this way is probably as cumbersome as you suggest. I can’t see how it wouldn’t be very messy. Perhaps too it inhibits the formulation of interesting disputes that actually take place among (at least some) scientists. That sounds right. Further I can’t see any scientific purpose to formulating such conditionals.
    But this I think is consistent with the central point that these conditions play a metaphysical role in determining what rule something is following. But then I’m not sure you disagree (or disagree entirely) with that.

  54. I’ve been trying to understand the original post. It seems to argue something like: (1) Fundamental epistemic principles are rules (of belief formation and change); (2) Only rule-following explains doxastic justification; (3) Rules followed (in belief formation and change) are actually local; (4) Fundamental epistemic principles are not local; therefore, (5) Local rules explain doxastic justification and (6) Fundamental epistemic principles do not explain doxastic justification (contradicting (1)); therefore, (7) Fundamental epistemic principles aren’t rules.

    Stripping it down, it looks like: (a) Suppose all principles are rules; (b) suppose all rules are local; (c) suppose all principles are nonlocal; (d) all principles are local (from (a) and (b)); and the only way this is not a contradiction is if there are no fundamental epistemic principles (so that (c) and (d) are not contradictory). Is that the argument?

    If I’m getting it right, then what’s left is to question whether principles are rules, which I take it Jon does question, that seeming to be rather the point of the original post; or that all rules are local, which I take it is being questioned in the multi-level rules discussion and, more generally, in the discussion of what counts as a rule; or that all principles are non-local, which I imagine we’re not going to question–these principles *are* supposed to be fundamental, after all. (Looking at the less stripped-down form, I see that we could also question whether only rule-following explains doxastic justification.) But I’ve had some trouble, I confess, fully digesting the original post; please pardon my denseness. Am I getting this right?

    Keith Brian Johnson

  55. Keith, all that is needed that is that some rules are local in cases where beliefs are doxastically justified, not that they all are local. Also, we don’t just suppose that true principles are nonlocal, that follows from the universality required for the truth of such principles. What follows from the locality of some rules is that true epistemic principles need not be fundamental, since the notion of fundamental being used here requires that only following rules that correspond to fundamental principles can yield doxastic justification.

  56. Pingback: Certain Doubts » Stats for August

  57. Pingback: Certain Doubts » Adam Morton on Rules and Principles

  58. Pingback: Certain Doubts » September Stats

Leave a Reply

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