Nice post! I’m not too in tune with this debate, but one thing I’m inclined to wonder about is the implications of the degree-of-support framing. In epistemology, we’d probably say that if you’re going to believe anything(i.e. accept it as true), it should be the thing you have most reason to believe—whatever it is that you have the highest credence in. We’d also say, for instance, that if there are two mutual exclusive propositions in which you have credences .3 and .7 respectively, you definitely *shouldn’t* believe the thing you have a lower credence in. If we map this onto a consequentialist morality, can’t we say something similar—that you should always do whatever you have most reason to do, and that if you’re choosing between two actions and you have more reason to do the former than the latter, you should do the former? To me, that seems like a roundabout way of reaching the conclusion that it’s obligatory to do whatever you have most reason to do. Curious what you think about this(and I hope I was sufficiently clear).
Thank you! I worry whether that’s smuggling in an extra discrete category over and above the finer grained “ground facts.” In the case of .3 vs .7 credence, I wouldn’t say that you should have that credence distribution and then there’s also some facts about what you should assent to. If you want to say that there’s this thing called “believing” over and above having a sufficiently high credence (determined by some vague, contextual threshold we decide), then I’d want to object! Once you’ve decided to have a credence of .7 in some proposition, we don’t then have to wait for you to also decide what to believe (putting it a little extremely).
Similarly, once we’ve said that you have such and such reasons to perform various actions, we don’t then have to decide what it is that you should do. What “should” is supposed to mean, according to me, is some amount of reason to take an action. You should take one of the actions that you have above X reason to take, because saying that you should do an action is just saying that you have above reason X to do it (again, there won’t be some deep fact about what X is, and it won’t be sharp).
Maybe I’m missing the force of your objection though?
Right, so I agree with you in the case of credences that, once you’ve figured out what your credence distribution is, you don’t have to take some further step of deciding what you actually believe. But maybe I can make my point clearer by discussing a case a case where you’re acting on the basis of your beliefs. You might know that if God exists, for instance, you’d want to do X, and that you’d want to do Y if God doesn’t exist. If you’re 70 percent sure that God exists, it seems like you should do X. To me, that case seems analogous to figuring out what to do on the basis of your moral reasons—once you’ve figured out your reasons distribution, you have to take the further step of deciding what you’re actually going to do. And when you’re making that decision, it seems right to say that—insofar as you’re capable—you should always do what you have the most reason to do. It strikes me that to do less would be failing to be appropriately sensitive to your reasons.
Edit: another way of putting the point—which is more appropriate for the way you’re using the word “should”—would be to ask why an appropriately reasons-sensitive agent would ever do anything other than what they have the most reason to do.
So in the God case, I don't think we have to appeal to beliefs at all; there we should just use some decision theory, and all we need is credences to figure out what to do.
But I think that might also be besides the point you're making, which seems to be more something like this (correct me if I'm wrong): when acting, we always have to do something. You can't partly take an action--either you do X or you don't. And so it seems that we have to also give some final answer as to what to do. You can't do 70% of the thing you have 70% reason to do, and 30% of the thing you have 30% reason to do.
I of course agree that you will always have to do exactly *one* thing. I just disagree that this shows that we have to put some special weight in deontic categories. Consider the agent who choses to do the thing that is slightly less than optimal. You're right that they're not *perfectly* reasons responsive. And so there's a sense in which they did the "wrong" thing. But they were only *slightly* unresponsive to reasons. What they did was *nearly* optimal. You can insist on saying that it's "wrong" to not be absolutely perfectly reasons-responsive. That's fine, but if we both agree on how strong the reasons for different options are, I just think you have a fancy way of talking--but it's not that someone who does the nearly optimal thing is doing something categorically different from the optimal person.
I think the problem is to always insist that we have to give some definite answer to what you "should" do. I don't think it's plausible that we can do this! Or at least, if we do this, we are smoothing over a lot of details.
Ok interesting, I think I’ve got a better grasp on where you’re coming from now. I’m still pretty uncomfortable with a moral system that doesn’t feature hard-and-fast obligations. That’s probably largely for psychological reasons, but I will say that standard deontic categories make it really easy to figure out how you should weigh your moral reasons against your other reasons, so to speak—a scalar conception of moral reasons might have more trouble with that issue, and I’d be curious how people who like scalar models deal with it. Setting that aside, though, I do now think my original objection fails. Scalar morality also becomes more intuitive, I think, when I compare succeeding to respond to moral reasons to success in other domains. Like if we consider an athlete—wrestling was my sport, so let’s say a wrestler—there’s a fact of the matter about whether she’s a good wrestler or whether she does a takedown in the right way, but the bar she has to clear to meet these standards isn’t fixed—it depends on how well other people do, and even the best wrestler in the world doesn’t do everything optimally. We might say that the concepts of right and wrong work in the same way. Whether it’s appropriate to apply these terms could depend on the context, but they’re not joint-carving(but the thing they propose to carve nonetheless exists). That makes a good bit of sense to me. So yeah, I guess I’m coming around to this scalar thing.
You say: "I will say that standard deontic categories make it really easy to figure out how you should weigh your moral reasons against your other reasons, so to speak—a scalar conception of moral reasons might have more trouble with that issue, and I’d be curious how people who like scalar models deal with it."
I'm not sure I see the issue there. Could you expand on that?
Yeah, absolutely(also, thanks for recommending that paper!). What I mean is that I have both moral reasons and self-interested reasons to act(at least—maybe I have epistemic reasons or something too). If the moral status of actions fits into a neat obligatory-permissible-forbidden structure, it’s really obvious how I should weigh my moral reasons against my other reasons—they take absolute priority in forbidden/obligatory actions, and otherwise I get to choose which reasons to prioritize(and it’s commendable if I choose the moral ones). But in a scalar system, it seems like you need to do some work to figure out how to weight your different reasons for action1. That isn’t a problem per se for scalar morality, it’s just an interesting issue it has to grapple with.
I feel like this is what happens when you put too much weight on theoretical simplicity as a virtue haha. We clearly need to be able to say things are right or wrong, obligatory or permissible, and it's so much more straightforward to just say there are deontic facts and it's not just what is convenient or helpful to say certain words. I think the same is true for the belief-credence case.
Obligations are how moral facts really get their oomph and explain substantive emotional states like blame and guilt, and this seems much easier with a simple deontic fact than with a comparative scalar fact which isn't fundamentally different between cases where someone did something bad but it was right (optimal) or equally bad but it was wrong (could have done better)
I just don’t find it that obvious that we need to say things are right or wrong, once we’ve settled how much reason we have to do certain things! Suppose we agree on how much reason you have to take various options, but not on what options are right or wrong. Do you think we substantively disagree about anything regarding what to do? It seems very obvious to me that once we’ve settled the facts regarding reasons, that’s all there is to say about what to do.
Regarding the case of blame and guilt: 1) Those things also seem to come in degrees, 2) you can say that those latch onto other facts than those regarding reasons (e.g. virtues/character).
What do you think of Spencer Case’s paper on scalar consequentialism and the demandingness objection? I like where Sinhababu argues that rightness and wrongness are actually scalar, and not deontic.
For Spencer Case, I just find it obvious that you have most reason to do the demanding things. I don’t find that to be an objectionable form of demandingness—I’d find it very strange if a theory got any other verdict!
Regarding what he says about guilt, it seems like it smuggles in a discrete category through the backdoor if the problem is deciding what you’re guilty for doing and what you’re not—you should just say that there are degrees of guilt. And it seems like guilt might also be accounted for through other normative categories than rightness, such as virtue or something like that. (I write about something like that here: https://wonderandaporia.substack.com/p/doing-and-allowing-for-utilitarians?utm_source=publication-search)
> but if we agree on the credences and disagree about which beliefs to have, we’re not actually in any substantive disagreement! We’re just talking past each other
What about agents who specifically want to act on their beliefs and not their credences? So, for example, suppose that there are two people who plan on lying to a police officer if they believe that the police officer will not arrest them for it. Let's posit that both of them have a credence of 60% that they will not get arrested for lying. But one of them believes that they will not get arrested and the other one doesn't. This means that one of them will lie and the other one won't. Clearly a substantial difference!
P.S. Otherwise I mostly agree with your argument and this is more of a nitpick than anything.
Nice post! I’m not too in tune with this debate, but one thing I’m inclined to wonder about is the implications of the degree-of-support framing. In epistemology, we’d probably say that if you’re going to believe anything(i.e. accept it as true), it should be the thing you have most reason to believe—whatever it is that you have the highest credence in. We’d also say, for instance, that if there are two mutual exclusive propositions in which you have credences .3 and .7 respectively, you definitely *shouldn’t* believe the thing you have a lower credence in. If we map this onto a consequentialist morality, can’t we say something similar—that you should always do whatever you have most reason to do, and that if you’re choosing between two actions and you have more reason to do the former than the latter, you should do the former? To me, that seems like a roundabout way of reaching the conclusion that it’s obligatory to do whatever you have most reason to do. Curious what you think about this(and I hope I was sufficiently clear).
Thank you! I worry whether that’s smuggling in an extra discrete category over and above the finer grained “ground facts.” In the case of .3 vs .7 credence, I wouldn’t say that you should have that credence distribution and then there’s also some facts about what you should assent to. If you want to say that there’s this thing called “believing” over and above having a sufficiently high credence (determined by some vague, contextual threshold we decide), then I’d want to object! Once you’ve decided to have a credence of .7 in some proposition, we don’t then have to wait for you to also decide what to believe (putting it a little extremely).
Similarly, once we’ve said that you have such and such reasons to perform various actions, we don’t then have to decide what it is that you should do. What “should” is supposed to mean, according to me, is some amount of reason to take an action. You should take one of the actions that you have above X reason to take, because saying that you should do an action is just saying that you have above reason X to do it (again, there won’t be some deep fact about what X is, and it won’t be sharp).
Maybe I’m missing the force of your objection though?
Right, so I agree with you in the case of credences that, once you’ve figured out what your credence distribution is, you don’t have to take some further step of deciding what you actually believe. But maybe I can make my point clearer by discussing a case a case where you’re acting on the basis of your beliefs. You might know that if God exists, for instance, you’d want to do X, and that you’d want to do Y if God doesn’t exist. If you’re 70 percent sure that God exists, it seems like you should do X. To me, that case seems analogous to figuring out what to do on the basis of your moral reasons—once you’ve figured out your reasons distribution, you have to take the further step of deciding what you’re actually going to do. And when you’re making that decision, it seems right to say that—insofar as you’re capable—you should always do what you have the most reason to do. It strikes me that to do less would be failing to be appropriately sensitive to your reasons.
Edit: another way of putting the point—which is more appropriate for the way you’re using the word “should”—would be to ask why an appropriately reasons-sensitive agent would ever do anything other than what they have the most reason to do.
So in the God case, I don't think we have to appeal to beliefs at all; there we should just use some decision theory, and all we need is credences to figure out what to do.
But I think that might also be besides the point you're making, which seems to be more something like this (correct me if I'm wrong): when acting, we always have to do something. You can't partly take an action--either you do X or you don't. And so it seems that we have to also give some final answer as to what to do. You can't do 70% of the thing you have 70% reason to do, and 30% of the thing you have 30% reason to do.
I of course agree that you will always have to do exactly *one* thing. I just disagree that this shows that we have to put some special weight in deontic categories. Consider the agent who choses to do the thing that is slightly less than optimal. You're right that they're not *perfectly* reasons responsive. And so there's a sense in which they did the "wrong" thing. But they were only *slightly* unresponsive to reasons. What they did was *nearly* optimal. You can insist on saying that it's "wrong" to not be absolutely perfectly reasons-responsive. That's fine, but if we both agree on how strong the reasons for different options are, I just think you have a fancy way of talking--but it's not that someone who does the nearly optimal thing is doing something categorically different from the optimal person.
I think the problem is to always insist that we have to give some definite answer to what you "should" do. I don't think it's plausible that we can do this! Or at least, if we do this, we are smoothing over a lot of details.
Ok interesting, I think I’ve got a better grasp on where you’re coming from now. I’m still pretty uncomfortable with a moral system that doesn’t feature hard-and-fast obligations. That’s probably largely for psychological reasons, but I will say that standard deontic categories make it really easy to figure out how you should weigh your moral reasons against your other reasons, so to speak—a scalar conception of moral reasons might have more trouble with that issue, and I’d be curious how people who like scalar models deal with it. Setting that aside, though, I do now think my original objection fails. Scalar morality also becomes more intuitive, I think, when I compare succeeding to respond to moral reasons to success in other domains. Like if we consider an athlete—wrestling was my sport, so let’s say a wrestler—there’s a fact of the matter about whether she’s a good wrestler or whether she does a takedown in the right way, but the bar she has to clear to meet these standards isn’t fixed—it depends on how well other people do, and even the best wrestler in the world doesn’t do everything optimally. We might say that the concepts of right and wrong work in the same way. Whether it’s appropriate to apply these terms could depend on the context, but they’re not joint-carving(but the thing they propose to carve nonetheless exists). That makes a good bit of sense to me. So yeah, I guess I’m coming around to this scalar thing.
Glad to hear you're becoming more sympathetic to the dark side!
If you're interested, I think this paper does a good job of intuitively motivating that scalar categories are a core part of moral evaluation: https://stafforini.com/works/sinhababu-2018-scalar-consequentialism-right/
You say: "I will say that standard deontic categories make it really easy to figure out how you should weigh your moral reasons against your other reasons, so to speak—a scalar conception of moral reasons might have more trouble with that issue, and I’d be curious how people who like scalar models deal with it."
I'm not sure I see the issue there. Could you expand on that?
Yeah, absolutely(also, thanks for recommending that paper!). What I mean is that I have both moral reasons and self-interested reasons to act(at least—maybe I have epistemic reasons or something too). If the moral status of actions fits into a neat obligatory-permissible-forbidden structure, it’s really obvious how I should weigh my moral reasons against my other reasons—they take absolute priority in forbidden/obligatory actions, and otherwise I get to choose which reasons to prioritize(and it’s commendable if I choose the moral ones). But in a scalar system, it seems like you need to do some work to figure out how to weight your different reasons for action1. That isn’t a problem per se for scalar morality, it’s just an interesting issue it has to grapple with.
I feel like this is what happens when you put too much weight on theoretical simplicity as a virtue haha. We clearly need to be able to say things are right or wrong, obligatory or permissible, and it's so much more straightforward to just say there are deontic facts and it's not just what is convenient or helpful to say certain words. I think the same is true for the belief-credence case.
Obligations are how moral facts really get their oomph and explain substantive emotional states like blame and guilt, and this seems much easier with a simple deontic fact than with a comparative scalar fact which isn't fundamentally different between cases where someone did something bad but it was right (optimal) or equally bad but it was wrong (could have done better)
I just don’t find it that obvious that we need to say things are right or wrong, once we’ve settled how much reason we have to do certain things! Suppose we agree on how much reason you have to take various options, but not on what options are right or wrong. Do you think we substantively disagree about anything regarding what to do? It seems very obvious to me that once we’ve settled the facts regarding reasons, that’s all there is to say about what to do.
Regarding the case of blame and guilt: 1) Those things also seem to come in degrees, 2) you can say that those latch onto other facts than those regarding reasons (e.g. virtues/character).
What do you think of Spencer Case’s paper on scalar consequentialism and the demandingness objection? I like where Sinhababu argues that rightness and wrongness are actually scalar, and not deontic.
For Spencer Case, I just find it obvious that you have most reason to do the demanding things. I don’t find that to be an objectionable form of demandingness—I’d find it very strange if a theory got any other verdict!
Regarding what he says about guilt, it seems like it smuggles in a discrete category through the backdoor if the problem is deciding what you’re guilty for doing and what you’re not—you should just say that there are degrees of guilt. And it seems like guilt might also be accounted for through other normative categories than rightness, such as virtue or something like that. (I write about something like that here: https://wonderandaporia.substack.com/p/doing-and-allowing-for-utilitarians?utm_source=publication-search)
I pretty much agree with Sinhababu
> but if we agree on the credences and disagree about which beliefs to have, we’re not actually in any substantive disagreement! We’re just talking past each other
What about agents who specifically want to act on their beliefs and not their credences? So, for example, suppose that there are two people who plan on lying to a police officer if they believe that the police officer will not arrest them for it. Let's posit that both of them have a credence of 60% that they will not get arrested for lying. But one of them believes that they will not get arrested and the other one doesn't. This means that one of them will lie and the other one won't. Clearly a substantial difference!
P.S. Otherwise I mostly agree with your argument and this is more of a nitpick than anything.