On Monday 28 Sept, we discussed:
We began by discussing a couple of proposed "Rationality Principles", which can be refined and interpreted in a variety of ways
One Means-End Principle says: If you intend end E, and (you believe/know that?) M is a necessary means to E, then you ought to intend M
One way of formulating a Closure Principle says: If you believe p, and (you believe/know that?) p entails q, then you ought to believe q
Enkratic Principles have the form: If you believe you ought to φ, then you ought to (intend to?) φ.
We can get theoretical/epistemic versions of this by substituting "believe p" in for φ. Broome insists that for practical enkratic principles, "intend to" has to be added in the consequent.
Notice that in all of these principles there are mental states specified in the antecedents that we're not requiring to be justified. The "narrow" interpretation of these principles sees them as having the form: X -> OY, and so merely having the states specified in the antecedent suffices for it being true that you ought to have the states specified in the consequent. Many judge that this predicts more oughts/obligations than intuitively exist.
The "wide" interpretation of the principles sees them as having the form O(X -> Y), and rejects the inference pattern of "factual detachment", so it would not follow merely from X that OY. Some who take this path are sympathetic to the inference pattern of "deontic detachment", which says that from O(X -> Y) and OX, OY does follow. But we will treat that as an open issue.
Intermediate interpretations are also possible, for instance: If you believe M is a necessary means to E, then O(you intend E -> you intend M).
How we evaluate these interpretive proposals will be affected by whether we think of the original principles as saying something like: X requires/commits you to Y; or whether we think of them as instead saying something like: a necessary condition for being "rational" is that if X, then Y. (Different notions of rationality may be substituted there.)
Two questions we will consider later are, first: How much of the wide-scopers' view essentially turns on their claim that the "ought" is capable of scoping differently wrt an independent conditional proposition, as opposed to: (a) the Kratzer-type view (advocated in Fogal's thesis) that the "if" or "given" clause expresses the (otherwise tacit) restrictor for a binary modal quantifier; or as opposed to: (b) the Schroeder-type view that "ought" doesn't take propositional arguments. Some parts of the wide-versus-narrow debate can be reconstructed in these alternative settings, but the terms in which we'd have to express the reconstructed views would look different.
Second, does the wide-versus-narrow debate need to be fundamentally an issue about strict requirements, or can the same sorts of issues also arise with more "slack" normative statuses, like pro tanto reasons.
We next turned to deontic logic, as surveyed here. We discussed the difficulty in interpreting O applied to sentences that don't describe you performing an action or making some optional choice. (For example, two of the axioms are O⊤ and P⊤.) At a later point in the discussion, the apparent difference in meaning between "It ought to be that you A" and "You ought to A" came up. We discussed the idea of interpreting Oφ to mean something like "A consequence of you performing all your obligations is that φ"; this is essentially the Kanger/Anderson strategy discussed in the survey. It makes the Rule RM (From ⊢ φ ⊃ ψ, infer ⊢ Oφ ⊃ Oψ) true of O just as a matter of meaning; some of the deontic puzzles challenged this inference rule.
I drew your attention to the interpretive issue illustrated in the survey document with the neighbor who has no business disciplining your kids, nor in preventing them from being disciplined. It's common to interpret Oφ, relative to a given agent N, as saying N ought to see to it that φ. But if the "see to it that" is part of the meaning of O, or part of the meaning of φ, then there are only two places to stick a negation. However we seem to want three. These seem to say three different things:
Some questions worth considering, but that we didn't go into in class: Is epistemic vocabulary like "reasonable," "justified," "warranted" well thought of in any deontic terms? If so, is it better thought of in terms of deontic necessity (mandates, obligations, requirements) or better thoght of in terms of permissions? Or perhaps in some kinds of intermediate terms? (Compare the "weak" necessity of recommendations like "should" to the "strong" necessity of "must"; is there such a thing as epistemic supererogation?) Or is epistemic vocabulary better thought of in terms that aren't deontic, yet are in some elusive sense still "normative" and not merely "evaluative"? Or is it just a kind of evaluation, perhaps evaluation in terms of conduciveness towards some end like truth or accuracy?
Does the choice of whether to interpret epistemic vocabulary in deontic terms interact in any interesting way with debates from the 1980s about the "deontological conception of justification"? (I don't think so, but perhaps there's something I'm not seeing.) Proponents of that conception of justification tended to invoke notions like "epistemic responsibility," and to unpack them in a way that involved highly reflective critical activity, of the sort we associate with Descartes' Meditations. They tended also to connect the notion of justification to notions of praise and blame. Their opponents complained that subjects could be blameless but still have defective (unjustified) beliefs. It's not clear what labels are appropriate for "anti-deontological" conceptions of justification. Some of those who oppose thinking of justification in "deontological" terms would reject the characterization that they understand it "merely evaluatively." Nor would they all want to embrace a "consequentialist" picture of epistemic normativity. In my view, this whole taxonomy never got sharply enough defined, though it was imaginatively productive in its day.
We discussed a couple of puzzles from the survey document challenging Rule RM. Ross's Paradox (O(mail the letter), so O(mail or burn the letter)) may raise different issues than the others. We then spent more time discussing Chisholm's Puzzle (O(visit), If you'll visit you ought to promise you will, If you won't visit you ought not to promise you will, You won't visit).
While discussing these puzzles, we introduced the idea of working with a conditional or dyadic deontic operator, O(φ | ψ). Plain O(φ) is then usually understood as O(φ | ⊤). As standardly axiomatized, we always have "deontic detachment": from O(ψ) and either O(ψ ⊃ φ) or O(φ | ψ), it follows that O(φ). Some want to handle Chisholm's Puzzle by rejecting this entailment.
At some point in this week's discussion, I drew attention to the fact that there are three communities of researchers: (a) those working on deontic logic, (b) those working on the semantics of ordinary language uses of "ought" and other deontic vocabulary, and (c) those working on considered (that is, not merely ordinary-language) normative theorizing. Sadly, there is only a little interaction between these communities. As a result, it's not unusual to see members of one community relying on naive, elementary positions from the domain of the other communities, and even to see them declare that their own work "solves" problems that the other community works on. (When in fact, the problem space has been developed to a much higher degree of sophistication than the ones making this declaration have familiarized themselves with.) We philosophers kind of suck.
I will tend myself to represent ordinary-language, conditioned "ought" claims using dyadic notation: O(not promise you'll visit | you won't visit), O(murder gently | you will murder), O(buy poison | you will murder, and poison is the only effective means). However I take the logic and semantics of these formulas to be a very open, debatable question. One common claim in the semantics for ordinary-language conditioned "ought" claims is that it's only in the worlds where the condition holds that the main, "prejacent" part of the ought has deontic necessity. We'll examine this claim later, for example when we read MacFarlane and Kolodny. My own view is skepticism that this is the proper interpretation of those formal semantic proposals. It's not clear that murdering gently and/or buying poison are actions that have any unconditional positive deontic status for you in any world, at least so far as statuses that we express in ordinary language go, or that play a role in serious normative theorizing. Their positive status seems, intuitively, to be essentially conditioned. Additionally, it's not clear that they have the same such status. So perhaps our formulas O(murder gently | you will murder), O(buy poison | you will murder, and poison is the only effective means) invoke different kinds of "ought".
Perhaps it would give us traction with Chisholm's Puzzle if we were able to distinguish the kind of normativity captured by "If you'll visit you ought to promise you will" from the "counter-to-duty" normativity captured by "If you won't visit (though you ought to), you ought not to promise". Positing two kinds of normativity is also motivated by some of the over-prediction problems facing "narrow" interpretations of the Rationality Principles we began with. (Though it's not at all obvious that the novel kind of normativity in both of these cases would be the same.)
There is a plethora of labels for these two kinds of normativity, none of them fully satisfactory. The more familiar kind of normativity gets called "reasons rationality" by Fogal, or evidential/justificatory norms by some (though "evidential" doesn't generalize well to the practical case). I've called it "categorical normativity", with the idea of opposing it to "hypothetical" or conditional normativity on the other side; but I didn't mean to prejudge issues about whether it's categorical in the Kantian or Footian senses (applying to agents regardless of their contingent properties). The more novel kind of normativity gets called "structural" and "attitudinal" rationality by Fogal, and as I said, I called it "hypothetical normativity." Broome initially called it "normative requirements" but then revised his terminology to "rational requirements" so as to not prejudge the question of whether it was genuinely normative (whether we had any reason to comply with these requirements). There is a broad tendency to use "rationality" in reference to this second category; though other authors use "rationality" to refer to the whole genus of which we're now distinguishing two varieties. Broome's use of "requirements" in the label prejudges some questions about the structure of the normative relations in this group (does it consist solely of requirements? or permissions? or "pressures", as Fogal argues?) Similar to Fogal's label of "structural rationality", some authors use terms like "coherence requirements" to label this second group. We'll see some questions that label prejudges in a moment.
What's common to the fans of hypothetical norms is that they characterize cases of normative goodness/badness that can hold between mere (not-necessarily-justified) attitudes. Also these fans reject the principle of factual detachment for these norms. The status of deontic detachment is contested. See also the questions raised at the end of the section on wide versus narrow scope, above.
Whether there is any interaction between the hypothetical and the categorical norms is an open issue, that different theorists have different views about. Fogal counts "reasons" and "structural" norms as two species of a single genus of "rationality", arguing that they have a unified internal structure. But he doesn't really make room for any kind of interesting agglomeration, interaction, or comparison between these species. (The only way to say which of two subjects is more rational overall is if they are equivalent in terms of one set of norms, but not in terms of the other.) Whereas my own view is that hypothetical norms and categorical norms can combine to generate more hypothetical norms.
I also think that hypothetical norms are irrelevant for questions of what you have prospective ("propositional") justification to believe or do, but they are important for questions of whether your belief or action is justified/well-founded. What I've found most interesting are questions not about hypothetical support but rather questions about hypothetical defeat, especially undermining.
In different ways, Fogal and I both think that hypothetical norms are something like "as if" categorical norms (they generate normative pressures that are "as if" you had some genuine justification for the attitudes they are hypothetical on, but these normative pressures are different in kind than would be generated by genuine justification).
Fogal opposes the "ideologies" that: (1) organize the phenomena around the notion of "strict rational requirements", and (2) give a foundational role to the notion of separate reasons. We won't be dealing much with his discussion of the second point, about the conflict between the balance of (mass-noun) reason and (count-noun) reasons, and how to understand the latter. With respect to the first point, Fogal advocates thinking of "slack", prima facie or pro tanto normative pressures as being explanatorily primary, instead of strict requirements. Here too my sympathies are aligned with his.
We discussed my complex version of the miner's puzzle, and how it's awkward to characterize the complex normative facts in these cases using just the pair of labels "subjective" and "objective". Also some genuine (not-merely-terminological) differences of opinion seemed to arise, about how other notions like advice and praise related to what's going on in the miner's case.
One reason I'm especially resistant of using the label "subjective" for the category of hypothetical norms is that it encourages the idea that the agent's beliefs about what she should do are determinative of the normative relations. (And perhaps they're all that's determinative.) But I think that idea is wrong: I agree the agent's beliefs about what she should do are contributory, but they don't trump other contributing factors, and we can find hypothetical normative relations also in cases where the agent has no higher-order opinions.
We discussed the question whether it's always "irrational" to be inconsistent. We could ask this question about both kinds of "rationality" (categorical or hypothetical); and instead of inconsistency, we could also ask about broader kinds of formal awkwardness, like probabilistic incoherence. Doesn't it seem like people can reasonably have false mathematical and logical beliefs? Perhaps the idea that inconsistency is always irrational fares better if we focus on known inconsistency. But the lottery puzzle is a familiar case where many think it can be reasonable for the agent to think of each ticket that it will lose, but reasonably think they won't all lose. Superficially, the Preface Paradox seems to have the same structure. But I pointed out that in the Preface case, your reasons for doubting that all of the beliefs in the book are true aren't (usually) coming from some evidence that opposes their conjunction. Rather, it's coming from some evidence that collectively undermines the individual beliefs (each to some small degree). This makes the Preface case interestingly different from the Lottery case. Many think the lesson of the Lottery and Preface puzzles is just: move to a graded notion of belief. But as we'll see, cases like the Preface can arise with graded belief too. So we should take it to be an open issue whether incoherence (perhaps even "known" incoherence) is always irrational or unreasonable or unjustified. More on this next week.
In the early part of Fogal's dissertation, he groups theorists into (a) those like Kolodny and Lord who only want to work with categorical norms, or as he says "reasons rationality"; (b) those like Broome who only want to work with hypothetical norms, or as he says "structural rationality"; (c) those like Parfit and Schroeder who are sort-of trying to reduce hypothetical/structural norms to facts expressed in terms of categorical/reasons norms (about which Parfit and Schroeder have very different views); and (d) pluralists like himself (and me) who think both kinds of norms exist and are interestingly different and in some ways independent.
Characterizing Broome as a "denier" of reasons rationality is somewhat inaccurate; he's more of an agnostic.
Kolodny and Lord are "deniers" of structural rationality in that, though they'd agree that alleged paradigms of structural irrationality really are normatively defective, they think this can be explained in terms of the agents' improper relations to their evidence or genuine practical reasons. Fogal gives a couple of arguments against Kolodny and Lord. I think the best of these are: (i) someone who is structurally defective seems to be exhibiting some further defect, beyond what we identified when we stipulated their beliefs didn't match their evidence (and indeed, sometimes the structural defect is because others of their beliefs do conform to their evidence); (ii) examples where you have higher-order evidence that opposes but doesn't swamp (even if it does interact with) your first-order evidence. In such cases arguably you could be justified in having first-order attitude A, but justified in believing that you aren't. Both your first-order attitude and your higher-order belief are justified; yet you seem to be manifesting some kind of structural badness (akrasia). Whether cases of reasonable akrasia really are possible is a controversy we'll be exploring.
Fogal also gives a number of other cases. These are all ones where there are some attitudes A1..An that seem individually, or in isolation, to be justified, but where you don't seem to justified in holding them jointly. For example, if we're pemissivists about some variety of justification, perhaps choice A and choice ¬A might be equally OK, but choosing both not be OK. Quinn's self-torturer (see also) gives another example of this. I agree that cases of hypothetical/structural badness exhibit this form; but I'm not convinced that everything exhibiting this form is a case of hypothetical/structural badness. I'm inclined to think there are just some facts about what combinations it's categorically/reasons appropriate to believe (or do or prefer or intend) that aren't a function of what's individually categorically/reasons appropriate. But I don't have an argument for this; nor do I know yet how to explain what makes the cases of hypothetical/structural badness distinctive.
Fogal discusses several variants of an example involving an agent Tom who has some settled but unjustified belief, yet is not sure whether to accept some obvious consequence of it. A variant agent Tim does accept the obvious consequence. (We suppose both the original belief plus the consequence are disconfirmed by these agents' actual evidence.) In a third variant, Ted continues to hold the original unjustied belief, but rejects its obvious consequence. About these cases, Fogal says: the third case Ted is in some ways more virtuous than Tom and Tim (because his beliefs better conform to his evidence), but there is some kind of virtue exhibited only by the second case, Tim. Tom and Ted's refusal to accept the obvious consequence exhbits some new, distinctive failing of theirs, additional to their failing in having the original unjustified belief. This despite the fact that, of all three, Tim's beliefs are the ones diverging most from his evidence.
Fogal also discusses two variants of an example involving an agent Jane who has some unreasonable end (counting blades of grass), yet refuses to do what she acknowledges to be a necessary mean to accomplish her end. In the second variant, Joan does carefully pursue that means. (She is therefore like Tim in the epistemic examples.)
In the epistemology literature, the dominant tendency is to emphasize categorical normativity, and thus if forced to rank these subjects, to rank Ted first. In the practical literature, the tendency is instead to emphasize coherence, and thus to rank Joan first. Fogal's view is that these agents all display a mixture of two orthogonal dimensions of virtue.