Epistemic Permissions and Epistemic Obligations

I’ve been thinking over the past few months about the notion of epistemic obligations and permissions, and will record a few thoughts here. There’s also a quite nice discussion going on at TAR about the subject, prompted by Roger White’s piece in the new Philosophical Perspectives.

Quite a few epistemologists think we have intellectual obligations regarding what to believe. I won’t object here to the idea that there are intellectual obligations, but what interests me more is the idea of epistemic permissions. In particular, I’m interested in the restrictive view that limits permissions to obligations, that (purely intellectual) permission can’t outrun obligation. Such a view would seem to be at a disadvantage, since the concepts of permission and obligation do not usually coincide. How fast you are legally permitted to drive is usually faster than the speed you are legally obligated to drive; your moral permissions vastly outstrip your moral obligations, etc.

What interests me about the restrictive view is what the view is, precisely, and what a theory of evidence will have to look like for the view to look plausible. Here I’ll comment on the first, and maybe get to the second point in another post. The goal I’m aiming for is to show that nobody holds the restrictive view.

So first a bit about the logic of the restrictive view. If we let P be an operator having to do with permissibility, O with obligatoriness, and F with forbiddenness, with psychological attitudes operators B and W (for belief and withholding) and a variable A ranging over these attitudes, the restrictive view wishes to endorse:
RV: P(Ap) iff O(Ap).

First, a bit of review of some axioms we’ll need in order to see whether a proposal counts as a version of the restrictive view (i.e., whether it is committed to RV). First, every view needs to start here:
Axiom 1a: P(Ap) iff ~F(Ap).
That is, permissibility needs to be understood in terms of failure of forbiddenness. One more that everyone should agree to:
Axiom 2a: O(Bp) iff F(B~p) & F(Wp).
That is, obligation to believe should be understood to involve the forbiddenness of disbelieving or withholding. To complete our explanation of the O operator, we can add:
Axiom 2b: O(Wp) iff F(Bp) & F(B~p).

Two final ones. First, if one attitude is forbidden, then one of the other two must be permissible:
Axiom 3b: F(Bp) only if P(B~p) v P(Wp).
Axiom 3b: F(Wp) only if P(Bp) v P(B~p).
(One interesting fact: if RV is true, then these ‘only ifs’ can be replaced by ‘iffs’.)

Finally, it can’t be permissible to both believe and disbelieve the same claim:
Axiom 4: ~(P(Bp) & P(B~p)).
(I have some doubts about this one, but it is so standardly assumed that I’ll use it below.)

The obvious way to endorse RV is to adopt the standard interpretation of the operators above. Doing so yields the claim that for every proposition, one must take one, and only one, of three attitudes toward it: belief, disbelief, or withholding. This is obvious enough, I think, but here’s a proof just for the fun of it.

Proof: Either the attitude is forbidden or it isn’t. If it isn’t, then it is permissible and hence obligatory. If it is forbidden, then there are only two attitudes left, and our proof requires three branches, depending on what is forbidden. Start with believing. If believing is forbidden, then what about disbelieving? If it isn’t forbidden, then it is permissible and hence obligatory; if it is forbidden, then only withholding remains and is obligatory by 2b. The argument when we start with disbelieving mirrors this one, so we can turn to withholding as the last possibility. If withholding isn’t forbidden, then it is permissible and hence obligatory. So it must be forbidden. But if withholding is forbidden, then either believing or disbelieving have to be permissible, by Axiom 3, and whichever one it is, it is obligatory. Hence, as advertised, the simple interpretation of the operators in RV requires that one hold one and only one attitude toward every proposition.

That view is, of course, crazy: it demands way too much of us to be plausible at all. A more plausible view is that for every proposition one considers, there is only one permitted attitude to take toward it. That’s a pretty restrictive view, but not one that identifies epistemic obligations and permissions. It leaves room for obligation and permission to come apart for propositions one has not considered.

Here’s something close to Roger White’s proposal: for every proposition, if you take any attitude toward it, there is only one permitted attitude to take (among those we’re discussing here). Is this view a restrictive view? Again, I think not. Take all the claims toward which you take no attitude. This view leaves open the possibility that permissions and obligations come apart here.

Notice the interpretations I’ve given of these last two proposals place the obligation operator in the consequent of a conditional. Perhaps a more charitable reading is to place the obligation operator out front, governing a conditional. But what conditional would it be? The antecedents are clear–either “propositions you consider” or “propositions you take an attitude toward”–but what is the consequent? Perhaps, for example, this: it ought to be the case that, if you take an attitude toward p then that attitude is the one and only attitude dictated by your total evidence.

There are two points to note about such a proposal. First, putting the obligation operator out front has a logical cost. Suppose you take an attitude toward p. If the obligation operator is only out front, we can’t infer anything about whether the attitude is appropriate. To do that, we need a principle with the operator on the consequent of a conditional.

Second, note the language of the consequent: the attitude dictated by the totality of one’s evidence. If there is such an attitude dictated in this way, then shouldn’t we be endorsing the simple account of RV above? Of course, we can’t, since that view places ridiculous demands on us. Yet, if we think the consequent of the conditional is necessarily false, we can’t use that conditional to explain the view in question.

In short, I doubt that anyone holds the restrictive view. Instead, some views are more restrictive than others. For example, what we might call experimentalists think that experience rarely speaks with singular clarity about what to believe, so that there is (almost) always rational optionality in one’s doxastic responses to experience. The alternative thinks of experience as placing more of a straightjacket on rational responses. How much more is where the action is, since the truly restrictive view–the view that collapses the logical distinction between epistemic permissions and epistemic obligations–is so hopelessly mistaken that no one should be tempted by it.


Comments

Epistemic Permissions and Epistemic Obligations — 18 Comments

  1. Jon, describing White’s view you say,

    “. . .if you take any attitude toward it, there is only one permitted attitude to take (among those we’re discussing here).”

    This conditional view has an odd consequence. Take any proposition p toward which you have taken no attitude. Are you permitted to take one? Well, simply as a matter of classical logic you are either permitted to take an attitude or you are not. Suppose you are. In that case, for at least some propositions p toward which you have taken no attitude, it is true that P(Ap). Of those propositions concerning which P(Ap), is it true of all that O(Ap)? Well, concerning those it is perfectly possible that O(Ap) iff.
    P(Ap). So even this conditional view is consistent with RV. It entails nothing either way about RV.

  2. “note the language of the consequent: the attitude dictated by the totality of one’s evidence. If there is such an attitude dictated in this way, then shouldn’t we be endorsing the simple account of RV above?”

    What if the consequent talks about the attitude *permitted* by the totality of one’s evidence? (NB talking about ‘the’ permitted attitude does not preclude the possibility that having no attitude is also permissible.)

  3. Carrie, that’s an interesting idea. It requires abandoning both versions of Axiom 2 above, I think, since otherwise an attitude could be obligatory and yet taking no attitude at all permissible. Then in any case where taking no attitude at all is permissible, there will be no epistemic obligations at all. That’s an interesting view, perhaps, but it isn’t a version of the restrictive view.

  4. Right Jon,

    But it is important that it does not weaken the restrictive view to adopt the conditional view. You need to adopt the conditional view and make other assumptions. So the problem with the restrictive view is not that it requires attitudes unconditionally. The source of the “restrictiveness” or demandingness lies elsewhere.

  5. Mike, I don’t understand this. The conditional view doesn’t entail the restrictive view, so if all you accept is the conditional view, you haven’t endorsed the restrictive view. You might, since the two are consistent, and if you already accept the restrictive view, you are committed to the conditional view. But neither of these points allows the conditional view to count as restrictive in the sense of collapsing the distinction between permission and obligation.

  6. Jon, suppose someone offers the unconditional view and, for the reasons you advance, we are persuaded that it is too restrictive. Someone says, “I have the remedy, it is this conditional view”. We say, “that isn’t the remedy, since that alone does not preclude the demandingness we’d like to avoid”. So, you haven’t really put your finger on the problem. At best it is the conditional view plus something else that solves the problem. But plus what?

  7. Jon, suppose someone offers the unconditional view and, for the reasons you advance, we are persuaded that it is too restrictive. Someone says, “I have the remedy, it is this conditional view”. We say, “that isn’t the remedy, since that alone does not preclude the demandingness we’d like to avoid”. So, you haven’t really put your finger on the solution. At best it is the conditional view plus something else that solves the problem. But plus what?

  8. Jon,

    “Then in any case where taking no attitude at all is permissible, there will be no epistemic obligations at all.”
    Won’t there still be epistemic obligations of the conditional form? If so, I think it’s restrictive enough to capture the spirit of the restrictive view.

    To get closer to the letter of RV, though, perhaps we can avoid the need for conditionals provided we only consider attitudes that people *actually have*? Say we have something like:

    For any subject S and any attitude S has to some proposition p, that attitude is epistemically permitted for S iff that attitude is epistemically obligatory given S’s situation.

    This avoids the need for the conditional since one relevant aspect of S’s situation is that she has got as far as adopting some attitude towards p.

  9. Jon,

    Looks like you are trying to derive RV from axioms (1a)-(4) since you say,

    “The obvious way to endorse RV is to adopt the standard interpretation of the operators above. Doing so yields the claim that for every proposition, one must take one, and only one, of three attitudes toward it: belief, disbelief, or withholding.”

    You proceed to the first part of the proof, which goes this way,

    “Proof: Either the attitude is forbidden or it isn’t. If it isn’t, then it is permissible and hence obligatory.”

    Consider belief. How do you get from P(Bp) to O(Bp) given those axioms? To derive O(Bp) you have to get F(B~p) & F(Wp). You can derive ~F(Bp) (fr. 1a) and ~P(B~p) (fr. 4). But if you do not assume RV in your proof, it is difficult to see how you arrive at O(Bp). If you do assume RV, then either the proof is circular or, more likely, you are not trying to derive RV. What am I missing?

  10. Mike, no, I wasn’t deriving RV from the axioms. What happened was that I wrote that RV says that you have to take one and only one attitude toward every proposition, and then started wondering if that claim really follows from RV. It’s pretty obvious that it does, but I formulated to axioms to see what was needed to get RV to imply the formulation in question.

  11. Carrie, we may be using the notion of a restrictive view differently. I’m thinking of it in terms of a collapse of the distinction between permissions and obligations. So if there are epistemic permissions without corresponding obligations, then the restrictive view is false.

    On your last formulation, that’s just RV, isn’t it?

  12. I suggested:

    RV’: For any subject S and any attitude S has to some proposition p, that attitude is epistemically permitted for S iff that attitude is epistemically obligatory given S’s situation.

    You asked: “that’s just RV, isn’t it?”

    I don’t think so. Suppose a subject S has no attitude to proposition p. RV says she’s permitted to believe p iff she’s obliged to believe p. RV’ does not have that consequence. RV’ has no consequences in cases where the subject has no attitude to the proposition, because of the quantifier phrases it begins with.

    That’s why RV’ doesn’t make the kind of unreasonable demands you’re worried about (demands like, for every proposition, our being obliged to have exactly one attitude towards that proposition). For all RV’ says, it could be that S is permitted but not obliged to *adopt* the belief that p.

    What RV’ does entail is that, once S has adopted an attitude towards p, the only permitted attitude will also be obligatory for someone in her situation (where her ‘situation’ includes the fact that she has some attitude towards p).

    RV’ isn’t RV but I would guess something like RV’ is what people have in mind when they talk about epistemic permission and obligation amounting to the same thing: the thought behind RV’ is that epistemic permission and obligation amount to the same thing as regards any attitudes someone actually has.

  13. Ah, I see now, Carrie. Here’s one of my concerns about the view, however. Any theory of permissions and obligations has to say something about attitudes that you don’t have. Your proposal can either align P’s and O’s there as well, in which case it is a version of the restrictive view. Or it might refuse to align P’s and O’s for attitudes you don’t have. In that case, it’s not a restrictive view. And it will have to explain the asymmetry between attitudes you have and attitudes you don’t.

  14. Yep, it’s not a restrictive view in the sense of being identical to RV, just an attempt to work out what people might really be up to when they say they want to align epistemic permissions and obligations.

    One way to explain the asymmetry you mention is to say that many (not all, I don’t think … but that’s another story) types of epistemic obligation only kick in once you get as far as adopting some attitude or other. There’s no obligation, for instance, to draw every permissible conclusion from your evidence. But once you’ve allowed your evidence to lead you to take some view on p, the only view you’re permitted to take is the one you’re obliged to take.

  15. Yes, Carrie, I agree that it’s a good account of what some aim for. I’m going to post on that view as well, but just wanted to start with something simpler.

    I tend to think like a Bayesian about these things, and first to think about what conditionalization will require when Bayesianism is given a human face. The first change will be to move to ranges of degrees, rather than precise degrees, I think. Once we do that, however, I think it’s hard to see how the obligation view can be maintained. We’ll have to identify coarse-grained attitudes with degrees in some way, and if we do, it’s hard to see why the range can’t bridge the line between different attitudes.

  16. Jon,

    Suppose an attitude “bridges the line” between attitudes. Say the agent does not definitely withold belief that p and does not definitely believe p. A defender of RV might reasonably say that no agent is required to take any indefinite attitude. But he could instead say that the agent is indefinitely required to believe that p. Or he could say that the agent is definitely required to indefinitely believe p iff. he definitely indefinitely believes p. And so on.
    For something closer to Bayesian degrees of belief it might be more convenient to adapt a degree theory. Someone who finds p exactly half-credible, believes to degree .5 that p and so (by some modified RV) is epistemically required to degree .5 to believe p. Nothing I can see precludes the development of subtler models along these lines for the obligation view.

  17. Pingback: Certain Doubts » White on Restrictivism and Optionalism

Leave a Reply

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