Print This Post Print This Post

In “Reasons as Evidence” and “Reasons: Explanations or Evidence?”, Daniel Star and I argue that a normative reason to A is evidence that one ought to A. In “Reasons”, John Broome argues that a normative reason to A is (part of) an explanation of why one ought to A (this characterisation is only rough). What both analyses have in common is that reasons are analysed in terms of oughts and not vice versa. I just wanted to ask what people think the prospects are for analysing oughts in terms of reasons. Here are a few natural ways to do this:

1) S ought to A iff S has most reason to A

2) S ought to A iff S has sufficient reason to A

3) S ought to A iff S has conclusive reason to A

4) S ought to A iff S has more reason to A than not to A

It seems to me that 1) won’t do because one may have MOST reason to do something (reply to an email, say), but still have sufficient reason to do something else at the same time (listen to some groovy music). It may thus be true that one does not have most reason to listen to some groovy music, but one still ought to do it.

Suggestions 2) and 3) have the following problem: the terms ‘sufficient’ and ‘conclusive’ are relative. Nothing is sufficient full stop or conclusive full stop. Something is sufficient FOR something else. What are sufficient reasons to A sufficient for? What are conclusive reasons to A conclusive about? It seems that the only answer available is that they are sufficient for (or conclusive about) the fact that one OUGHT to A. But this means 2) and 3) are not really analyses of oughts.

Suggestion 4) seems most promising, but there are some problems with it. Most obviously–what if S has only just more reason to A than not to A? Isn’t it then possible that it is not the case that S ought to A? (The natural response to this problem is to distinguish between oughts and obligations, but I’m not sure this distinction can do all the work it needs to here). A related problem with 4) is that it conflicts with some ways of spelling the idea of superogatory actions.

Another problem with 4 is that it invokes the notion ‘more reason’, which presumably should be spelled out in terms of the reasons to A outweighing the reasons not to A. But this notion is difficult to understand without some analysis of it. Star and I give such analysis (or at least the outlines of one) in our papers, but this is not available to a proponent of 4).

Perhaps there are some other obvious accounts that avoids these problems.


Comments

  1. 1. Posted by Geoffrey Ferrari | February 12, 2008 12:41 pm

    Hi Stephen,

    Thanks for this interesting post. I’m currently thinking about the related issue of whether its possible to analyse “making” relations such as rightmaking and wrongmaking in terms of reasons. If you’re looking for a specific opponent, I think Dancy in “Ethics without Principles” argues that reasons cannot be analysed in terms of overall oughts. However, he doesn’t consider any analysis along yours or Broome’s lines.

    I’ve got some comments on the particular formulations you mention.

    First, some clarificatory questions:

    (i) which sense of “ought” are you trying to analyse? I take it it’s the overall, “all things considered” (ATC) sense.

    (ii) Do you accept the general principle that if you ought ATC to do an act A then it is ATC impermissible/wrong not to do A?

    (iii) Do you accept that ought ATC implies can? That is, can it be the case that one ought ATC to do A, ought ATC to do B, but that one cannot do both A and B?

    (iv) Do you accept the agglomeration principle that ought A and ought B implies ought (A and B)?

    Now let me say why I think these questions are important.

    You argue against your first analysis, (1), that one may have most reason to do one action, call it A, and yet have sufficient reason to do another action, call it B, so that it can still be the case that one ought to do B.

    I read your objection here as denying that having most reason to do A is necessary for it’s being the case that one ought to do A. That is, you’re denying that ought implies having most reason. This raises several questions.

    First, in the example you give, one has most reason to do A, yet one ought to do B. My first question is, ought one also to do A? If one ought to do A, and one also ought to do B, then by the agglomeration principle one ought to do (A and B). Yet (so I assume for your example) one cannot do (A and B). So if in your example you think you ought to do A as well as B, then you need to deny either the agglomeration principle or the ATC ought implies can principle. One might think these are high prices to pay.

    The alternative for you is to deny that one ought to do A in your example. Yet one has most reason to do A. So now you need to deny that not only is having most reason not nececessary for an ought, it is not sufficient either. One has most reason to do A, yet that it is not the case that one ought to do A.

    So to sustain your first example, you’re faced with a dilemma. You either need to reject both directions of (1), or, if you want to deny only that ought implies most reason, you need to reject one of two principles, either the agglomeration principle or ATC ought implies can principle. Which of these options do you prefer?

    You object against your second and third analysis, (2) and (3), with the claim that “nothing is sufficient full stop or conclusive full stop”. Instead, you suggest that we must mean sufficient for or conclusive about the fact that one ought to A. I think I see your point, and you might support it further by noting that Parfit, in his Climbing the Mountain manuscript, takes having sufficient reason and having conclusive reason as distinguishable senses of, precisely, “ought”.

    I think in the end this is a good objection to these analyses. EITHER one has to analyse sufficient and conclusive reason as “reasons that imply an ought”, or else one to provide another analysis or take the notions of sufficient and conclusive reason to be basic. The only other analysis I can think of is in terms of permissibility: one has sufficient reason to do A iff it is permissible that one does A. But there seem to be cases of permissible acts that one has no good reason to do (e.g. touching one’s nose with one’s tongue), so it would stretch things to say one has sufficient reason to do them! Also, an analysis in terms of permissibility wouldn’t work for conclusive reasons. It seems you’d need something stronger than permissibility for that. Finally, of course, an analysis in terms of permissibility wouldn’t really provide an analysis in terms of reasons.

    So the only remaining option is to take the notions of sufficient/conclusive reason to be basic. This seems like a poor option to me, but I find it hard to say why. Perhaps these just don’t seem like basic notions.

    Finally, a brief word on your fourth analysis. I accept your objection that sometimes the reasons are too finely balanced. So how about this: (4*) S ought to A iff S has significantly more reason to A than not to A. Attention here will surely focus on “significantly”, but one possibility is that we should let this be determined contextually. As a result, we’ll find that ATC ought is also a contextual notion. But that is something one might be willing to live with.

  2. 2. Posted by Doug Portmore | February 12, 2008 10:14 pm

    Stephen,

    You write:

    Suggestions 2) and 3) have the following problem: the terms ’sufficient’ and ‘conclusive’ are relative. Nothing is sufficient full stop or conclusive full stop. Something is sufficient FOR something else. What are sufficient reasons to A sufficient for? What are conclusive reasons to A conclusive about? It seems that the only answer available is that they are sufficient for (or conclusive about) the fact that one OUGHT to A. But this means 2) and 3) are not really analyses of oughts.

    Doesn’t this assume that the most fundamental normative notion has to be either ‘ought’ or ‘reason’? But couldn’t it be that neither of these two are primitive and that instead the only primitive normative notions are those pertaining to normative status: notions such as ‘required’, ‘permissible’, ‘wrong’, etc. (This might be Josh Gert’s view; he does, at least, think that ‘reason’ isn’t primitive and is to be analyzed in terms of normative status.) If this is right, one could deny that “the only answer available is that they are sufficient for (or conclusive about) the fact that one OUGHT to A.” The other answer is that sufficient reasons are sufficient for generating a permission, and that conclusive reasons are conclusive with respect to generating a requirement. And, in turn, you could analyze ‘ought’ in terms of ‘reason’.

  3. 3. Posted by Stephen Kearns | February 13, 2008 3:25 am

    Thanks Geoff and Doug for your great comments,

    Geoff: With regards to 1), I was actually thinking that A and B could both be done. Thus one has MOST reason to write an email but sufficient reason to ALSO listen to some music. Thus one ought to do both (and, indeed, one CAN do both). This means I can avoid having to deny those plausible principles you bring up. Thus I don’t need to deny that having most reason implies ought. It’s rather that ought does not imply having most reason.

    And, yes, I was thinking of ‘ought, all things considered’, though I’m not sure that there’s any other type. As to what I think about ought implies can, I’m tempted to ask: which ‘can’? It is metaphysically impossible for me to A and B? Physically impossible? Something else?

    On 2) and 3) you say that maybe ‘sufficient’ could mean ‘sufficient for permissability’. I agree–it could mean that. But in that case, the analysis of ought (One ought to A iff one has sufficient reason to A) would clearly be wrong! Also, permissability is usually defined in terms of oughts (one is permitted to A iff it is not the case that one ought not to A).
    I also don’t think taking ‘sufficient’ and ‘conclusive’ as primitive will work. The only way it would is if we take these terms to be terms of art, but then we don’t really have an analysis.

    On 4), I’m a bit wary of a contextualism about ‘ought’. When I’m figuring out what to do, it doesn’t help to know that, in one context ‘I ought (ATC) to A’ expresses a truth and in another context it expresses a falsehood. (see Williamson’s ‘Knowledge, Context and the Agent’s Point of View’ for a similar point).

    Doug: I guess I was wondering in the post whether oughts could be analysed in terms of reasons, not whether both could be analysed in other terms. Still, I find it hard to see something like ‘S ought to A iff S is normatively required to A’ as an analysis of ought. This so for three reasons. One is that ‘normatively requires’ is a term of art. Another is that it strikes me that one can only have the concept of a normative requirement if one ALREADY has the concept of ought. The other is that they may just mean the same thing as each other. In which case, we’re not really analysing one in terms of the other. So I agree with you that one may appeal to normative requirements to analyse what ‘conclusive reason’ might mean, but this doesn’t help us with an analysis of ‘ought’.

    Thanks again for your comments. I hope my response is relatively clear.

  4. 4. Posted by Doug Portmore | February 14, 2008 7:36 pm

    Stephen,

    You’re wondering whether oughts can be analyzed in terms of reasons. One proposal is:

    3) S ought to A iff S has conclusive reason to A.

    You rejected 3, because it seemed to you that ‘conclusive reason’ can be analyzed only in terms of ‘ought’. Now you concede that we might be able to analyze ‘conclusive reason’ in terms of ‘required’. Perhaps, this is a suitable analysis:

    S has conclusive reason to A =def. S is required to A. (Note that there’s no need to employ any technical notions here.)

    So what now is your objection to 3? I gather that it is that you think that ‘required’ can only be analyzed in terms of ‘ought’. But someone else might claim that it is the other way around: ‘ought’ can only be analyzed in terms of notions such as ‘required’, ‘permissible’, and ‘impermissible,’ for it is these latter notions and not ‘ought’ that are primitive. So aren’t you just assuming that ‘ought’ is more fundamental than these other notions? But my question is: Why not think that notions such as ‘required’, ‘permissible’, and ‘impermissible’ are more fundamental or, indeed, primitive? Don’t get me wrong. I’m not claiming that these notions are primitive notions; I’m trying to understand your reasons for thinking that ‘ought’ cannot be analyzed in terms of ‘reasons’ and whether there’s some assumption here that a notion like ‘required’ isn’t primitive and, if so, what your reasons are for thinking this assumption is true.

  5. 5. Posted by Stephen Kearns | February 16, 2008 7:36 am

    Doug,

    I have a few points in response:

    I conceded that we might be able to analyze conclusive reasons in terms of normative requirements (though, given my commitment to analysing them in terms of evidence of oughts, I would oppose such an analysis), but I don’t think one could plausibly *define* ‘conclusive reason’ as ‘required’. The phrase ‘conclusive reason’ is clearly not simple and it’s meaning is a function of the meanings of ‘conclusive’ and ‘reason’. Given this, in order to define ‘conclusive reason’, we’d first need a definition of ‘conclusive’ and of ‘reason’.

    If conclusive reasons are merely analysed in terms of normative requirements (and not defined as such), then analysing oughts also in terms of requirements does not give us an analysis of oughts in terms of reasons.

    Secondly, I’m not optimistic of an analysis of oughts in terms of requirements. As I understand them requirements are three place relations–X requires that Y do Z. Oughts are two place. Y ought to do Z. I think that perhaps requirements may be analysed in terms of wide-scope oughts (Y ought, if X, to Z), but I’m not sure how one might do it the other way around. (Saying that Y ought to do Z iff there is something that requires Y to do Z seems wrong because, as Broome points out, sometimes we are required to do something by a belief or intention that we in fact ought to not do.)

    I’m not sure about any of this, but my suspicion is that oughts are fundamental and either requirements are too, or they can be analysed in terms of oughts. Even if requirements are primitive and oughts can be analysed in terms of them, this does not amount to an analyse of oughts in terms of reasons.

  6. 6. Posted by Clayton Littlejohn | February 17, 2008 4:49 pm

    I’m somewhat surprised that there wasn’t:

    (5) S ought to A iff there is an undefeated reason for S to A.

    So long as we can explain the notion of defeat without using the notion of ought, (5) seems more promising than (1)-(4). I think those attracted to the sufficiency account or the conclusive reason account should be attracted to (5) insofar as it gives us a natural way of saying what it is for a reason to be sufficient or conclusive. A reason is sufficient in a set of circumstances when it is unopposed or overrides the other reasons present in those circumstances. The same goes for conclusive.

    I’m sure there are problems with (5), but it should at the very least be on the dustbin (in the dustbin? I guess it’s getting full.)

  7. 7. Posted by Stephen Kearns | February 21, 2008 7:09 am

    Clayton,

    You’re quite right that (5) should be a contender. I like your suggestion also that it gives us a nice way of understanding sufficient and conclusive reasons.

    For someone who thinks oughts are more fundamental, there are two possible ways to attack (5). First, claim that (5) is false. Second, claim that (5), even if true, is not really an analysis.

    Here is a reason to think (5) is false. It seems that if Aing and not Aing are both permissable, there may be undefeated reasons to A and undefeated reasons not to A (these reasons aren’t overridden after all). Thus it is possible for there to be undefeated reasons to A even when it is not the case that one ought to A.

    As for whether (5) is really an analysis, this depends (as you suggest) on whether the notion of an undefeated reason (and the related notions of being unopposed or being overriding) can be spelled out without appealing to oughts. It seems that these notions are relative in two ways.

    First, one may ask questions such as ‘Undefeated by what?’; ‘Unopposed by what?’ and ‘Overrides what?’. The answer to all these questions is: other reasons. This is fine so far.

    Second, we may ask ‘Undefeated relative to what goal?’, ‘Unopposed relative to what goal?’, etc. It is here I suspect that one will need to appeal to oughts in order for (5) to be true. In which case, (5) will not be an analysis. Obviously, that’s just a suspicion as it stands, but a natural one to have. There’s obviously a lot more to say and probably many more possible analyses of oughts to consider (perhaps, S ought to A iff there a winning reason to A?).

  8. 8. Posted by Clayton Littlejohn | February 22, 2008 4:12 pm

    Hey Stephen,

    I think the winning reason version does a nice job handling the case you thought showed (5) to be false. I think that like you I’m worried that at some point we won’t be able to tell our story about defeat (or winning) without helping ourselves to claims about oughts. I’m not entirely sure we need to specify goals relative to which one reason defeats another. My worry is that the natural candidate for spelling out what it is for one reason to defeat another is in terms of comparative weight or something of the sort and that’s just not going to do justice to the fact that the inference from claims about reasons to claims about ought doesn’t rest entirely on claims about comparative weights. Exclusionary reasons have to enter into the mix at some point and those are reasons that do their work while ‘punching above their weight’ as I think someone once put it.

  9. 9. Posted by Geoffrey Ferrari | February 25, 2008 1:55 am

    Dear Stephen et al,

    I have a few further comments to add to the discussion.

    Regarding the proposal to analyse oughts in terms of normative requirements. Broome argues against this in his “Reasons” paper. The essential problem is that (as Broome supposes) normative requirements take only wide scope, whereas oughts do not. That is to say, for example, one may be normatively required not to (believe P and believe not P), or normatively required to (believe Q if you believe both P and if P then Q). But, if I understand him correctly, Broome would not accept that one can be normatively required to P full stop (where P is an atomic proposition). Nevertheless, cases of oughts governing atomic propositions seem fine. So oughts cannot analysed in terms of normative requirements.

    Let me return to Stephen’s original suggestions.

    Regarding the claim that one ought to A iff there is sufficient reason to A. You claim (albeit tenatively) that “sufficient” can only be analysed here in terms of being sufficient for an ought. But compare satisficing consequentialism: an act is right iff it is sufficiently good. Here “sufficiently good” is never (to my knowledge) thought to mean “good enough to ground rightness”. It may be, for example, sufficient to raise goodness above some absolute or relative threshold. So it may be possible to understand sufficient reason along similar lines. Here’s a concrete example: In “Reasons and The Good” Crisp suggests that ultimate reasons are grounded in welfare. So perhaps: one ought to A iff there is sufficient reason to do A iff A will raise the total welfare of sentient being above some absolute or relative threshold. A similar line of thought may or not work for “conclusive reason”: e.g. there is conclusive reason to A iff A will raise the level of total welfare to some conclusive threshold. But this seems less defensible as it stands.

    Regarding the claim that one ought to A iff there is most reason to A. You objected against this, Stephen, with an example in which one has most reason to A but sufficient reason to B, and you later clarified that in this example, it is possible to do both A and B. In that case, I don’t see why you mention B at all. I could understand your objection if doing B implied not doing A. For in the case where one can do only one of A and B, there is a good explanation why it is not the case that one ought to A, even though there is most reason to A, namely, that one has sufficient reason to B, and one can do only one of A and B. But in your case, where A and B are both possible, there is no such explanation available. The case simply amounts to one where one has most reason to A but it is not the case that one ought to A. It just seems like a brute assertion that such cases are possible.

    For further discussion, let’s consider a case like yours, but where A and B are mutually exclusive. One has most reason to do A, but it is not the case that one ought to do A. There is sufficient reason to B, and one ought to B. And A and B are mutually exclusive. Could there be a case like this? Obviously, such cases imply the denial of reasons maximisation (the claim that one ought to A iff there is most reason to A). But they also imply the denial of the further claim that there is sufficient reason to A iff there is most reason to A. (Here I should make clear that I am using “most reason” in the sense of “having no alternative for which there is more reason”, to cope with cases of incomparability and ties for first place.) Is is plausible to deny that most reason and sufficient reason are equivalent? Let me give one story why not. As Broome points out, talk of reasons often proceeds by analogy with mechanistic weighing. Can you think of a sense of “sufficient” in which there is most weight to one side of the balance but not sufficient weight on that side? Or vice versa? I can’t, so I’m not sure there could be a sense of “sufficient reason”” that isn’t equivalent to “most reason”.

    Finally, I’ll remind you of one further point in Broome’s paper. Broome claims that not all oughts are explained by reasons. For example, a side-constraint on killing may ground the claim that one ought not to kill directly, without providing reasons for it. Broome supports this claim with the idea that reasons, pro tanto reasons, that is, are the kinds of things that feature in weighing explanations, whereas side-constraint principles don’t work like that. If you accept this, one could deny both directions of any analysis of oughts in terms of pro tanto reasons. Side-constraints can ground oughts without providing pro tanto reasons. And there can be cases where any amount of pro-tanto reasons would not ground an ought because the act in question is ruled out by a side-constraint.

  10. 10. Posted by Geoffrey Ferrari | February 26, 2008 4:32 pm

    You may be interested to know that in “The Nature of Morality”, p. 118, Harman discusses something close to the following analysis:

    S ought to do A iff S has sufficient reasons to do A that are stronger than the reasons S has to do anything else

    Or, equivalently (I presume):

    S ought to do A iff [S has sufficient reasons to do A and S has most reason to do A]

Post a comment

Name: (required)

Email Address: (required) (will not be published)

URL:

Comments: