Abstract
There are plenty of classic paradoxes about conditional obligations, like the duty to be gentle if one is to murder, and about “supererogatory” deeds beyond the call of duty. But little has been said about the intersection of these topics. We develop the first general account of conditional supererogation, with the power to solve familiar puzzles as well as several that we introduce. Our account, moreover, flows from two familiar ideas: that conditionals restrict quantification and that supererogation emerges from a clash between justifying and requiring reasons.
Similar content being viewed by others
1 Introduction
The point of morality—if it has one—is to guide our actions. Moral guidance usually comes in the form of obligations, which steer us away from wrongs like theft and murder. But wrongdoers need guidance, too. For their sake, morality issues conditional obligations. For example: if you are going to murder, you must do it gently. What makes this obligation “conditional” is that it applies given a certain condition—your murdering. What makes it “contrary-to-duty” is that the condition is a wrong action (Chisholm, 1963). There are also what we call “consistent-with-duty” conditional obligations, like the obligation to wear a seatbelt if driving. These provide guidance to those who have ruled out a non-obligatory option, like taking the train.
The study of conditional obligations has been animated by a stock of paradoxes. The “gentle murder” paradox, for example, involves a tempting inference known as “factual detachment.” If you are going to murder someone, then you must do it gently. In fact, you are going to murder. Does that mean you must commit a gentle murder? Clearly, this cannot follow (Forrester, 1984; cf. Jackson, 1985, pp. 191–92; McNamara, 2019). The classic task is to explain why.
Contemporary work on conditional obligations has been in fruitful dialogue with the flourishing study of conditionals (see e.g., Smith, 1993; Bonevac, 1998; McNamara, 2010).Footnote 1 Also flourishing is the literature on obligation—both as a topic in itself, and especially in relation to “supererogatory” acts that lie beyond its call. But this work has not been put in contact with the inquiry into conditional moral judgments. Our task in this paper is to fill this lacuna, seeking an account of conditional judgments fit for the supererogatory, and laying out some hard cases with which any such account must contend.Footnote 2
We argue that every existing principle struggles with at least some of the hard cases: either consistent-with-duty conditional obligation (Sect. 2), consistent-with-duty conditional supererogation (Sect. 3), or contrary-to-duty conditional supererogation (Sect. 4). We then develop a principle that can handle them all. Our principle combines a familiar conception of conditionals (as restrictors on quantification) with a key resource from the theory of supererogation—namely, the idea that permissibility depends on the balance of “requiring reasons” and “justifying reasons” (Sect. 5).
2 From obligation to conditional obligation
Let’s start by thinking about conditional obligations in simple choices involving supererogation. A supererogatory act, like a friendly favor or saintly sacrifice, is permissible and yet better than a permissible alternative—it goes “beyond the call of duty.” The supererogatory contrasts with the “moral minimum,” i.e., the minimally decent permissible option (McNamara, 1996).Footnote 3
We begin with a three-way choice between supererogating, doing the moral minimum, and doing wrong. Suppose that two people, both strangers to you, are trapped in a collapsing building, and while you can easily and costlessly save one of them, saving both would involve serious harm to you—say, losing your legs. You thus have three options: Save Two (supererogatory), Save One (less good but still permissible), and Save Zero (wrong). Clearly, you aren’t obligated to Save One, since you may instead Save Two. But if you won’t Save Two, then you must Save One. It would be wrong to gratuitously let a stranger die.
We can capture this with a bridge principle between non-conditional and conditional obligations. (We write ‘OB(B/¬A)’ to mean that B is obligatory conditional on not doing A.)
Nothing Else Left: If A and B are your only permissible options, OB(B/¬A).Footnote 4
The idea is that you are conditionally obligated to do something if, given the conditions, it is the only (non-conditionally) permissible option left.Footnote 5
But as nice as it sounds, Nothing Else Left has trouble with certain consistent-with-duty conditionals (the principle is silent about contrary-to-duty conditionals). Consider a case known for giving rise to the “All or Nothing Problem” (Horton, 2017).Footnote 6 We have another collapsing building with two strangers trapped inside, but this time you’ll lose your legs whether you save one stranger or save both—saving zero remains costless. You have three options: Save Zero (no cost), Save One (at the cost of your legs), or Save Two (at the cost of your legs). In this case you are plausibly permitted to Save Zero, as this is the only way to keep your legs, and you are obligated not to Save One, because Save Two is far better and no costlier. Again, it’s wrong to gratuitously let a stranger die.Footnote 7
Those are your obligations. What about your conditional obligations? Here is where the problem starts. If you won’t Save Two, it seems you have only one permissible option left—Save Zero. But does it follow that, if you won’t save everyone, you must save none (rather than saving one)? That seems perverse. This is the All or Nothing Problem, and its core is Nothing Else Left. It is true that you have only two permissible options: saving “all” or saving “nothing,” but it shouldn’t follow that you must choose nothing if you won’t choose all.Footnote 8
3 Is “next best” good enough?
To solve the All or Nothing Problem, we will need a new kind of principle. We now turn to alternatives inspired by work on dyadic conditional obligations (Comesaña, 2015; Hansson, 1969; Lewis, 1973). Here is the simplest. (We write ‘PE(B/¬A)’ to mean that B is permissible conditional on not doing A; ‘iff’ means ‘if and only if’.)Footnote 9
Next Best: PE(B/¬A) iff B is the best option compatible with ¬A.
If you won’t do A, you may do the best option left on the menu once A is removed. (And you are conditionally obligated to choose from your conditionally permissible options.)Footnote 10
A perk of Next Best is that it can give advice to agents who will act contrary to duty. If you will murder, you must murder gently, because even though any murder is worse than no murder, gentle murder is the next best option (assuming your only options are murdering brutally, murdering gently, and refraining from murder altogether). But Next Best, like Nothing Else Left, struggles with the consistent-with-duty permissions of the All or Nothing Problem. Next Best implies that, if you will not Save Two, you must Save One. This seems too demanding. If there were just one person in the building, you would not have to sacrifice your legs to save them. This sacrifice would be supererogatory. So why should that same tradeoff be obligatory, conditional on not saving both strangers in the All or Nothing Problem? It ought to be conditionally optional.
This leads Pummer (2019, p. 286) to a revised principle, which like Nothing Else Left infers conditional permissibility from plain old non-conditional permissibility:
Okay or Next Best: PE(B/¬A) iff (i) PE(B), or (ii) B is the best option compatible with ¬A.Footnote 11
But if an option is wrong only because it is worse than A, then it is permissible conditional on not doing A.
Okay or Next Best, like Nothing Else Left, gets the right answer in the simple supererogation case. If you won’t Save Two (at the cost of your legs), then you must Save One (no cost). That is the next best option, and the only one that’s non-conditionally permissible.
Moreover, with Okay or Next Best, we can solve the All or Nothing Problem. Conditional on not saving two, you may save either one or zero. Because Save Zero is permissible, it is also conditionally permissible. Save One, meanwhile, is conditionally permissible for a different reason: it is the next best option, given that you won’t Save Two. Okay or Next Best thus implies that saving the one stranger is an instance of conditional permission—indeed, of consistent-with-duty conditional supererogation.Footnote 12 In this sort of case, Okay or Next Best performs better than Nothing Else Left.
You might wonder, “Who cares about conditional permissions? What kind of guidance do we get from hearing that an option is conditionally permissible?” Of course, we don’t get the same direct guidance—“do this!”—that we would get from obligations. But again, this guidance isn’t for everyone. Wrongdoers, like the would-be gentle murderer, need the guidance of conditional obligations as a supplement. Once they have excluded not murdering, they need to be guided away from brutal murder towards gentle murder. Non-supererogators, meanwhile, need conditional permissions as a replacement, a way of getting up-to-date advice as they whittle down their option sets over the course of deliberation. At the outset, morality warns against Save One—it’s decisively worse than Save Two. But that advice is inapt when offered to someone who has already excluded Save Two from deliberation, because it would guide them away from Save One towards Save Zero. You shouldn't have to save nobody if not everybody. That’s why—even though Save One is plain wrong—it has to be permissible conditional on not saving two. Conditional permissions are crucial: they can preempt the disastrous imperative to move from second-best to very worst.Footnote 13
4 The hardest case: contrary to duty, beyond the call
We have seen that there are counterexamples to Nothing Else Left in cases of wrong but conditionally permissible actions. Are there any counterexamples to Okay or Next Best? We think there are, in cases of conditional supererogation.
First, consider Kamm’s (1985) influential example from “Supererogation and Obligation.” You have three options: you can Keep Your Promise to meet a friend for lunch (permissible), break your promise to Do Nothing at home (wrong), or break your promise to Save One stranger’s life at great cost to yourself (supererogatory). Save One is the best option compatible with excluding Keep Your Promise. So, according to Okay or Next Best, if you won’t Keep Your Promise, you must Save One, which costs you your legs.Footnote 14 But this is not plausible, as Kamm herself observes.Footnote 15 After all, even though you can’t justify Do Nothing over Keep Your Promise, since there is a negligible difference in costs to you, it is much costlier to Save One, and so it seems you may invoke the costs of heroism to justify your choice to Do Nothing instead.Footnote 16
In Kamm’s case (which has not been widely discussed in the context of conditional obligations), there is a conflict between supererogation and what would ordinarily be an obligation—viz., Keep Your Promise.Footnote 17 It is supererogatory to Save One, and indeed it should remain supererogatory even if you are not going to keep the promise. Life-saving is thus conditionally and non-conditionally supererogatory.Footnote 18
So why does Okay or Next Best struggle to get the right result in Kamm’s case? Perhaps the problem is that the next best option if you won’t Keep Your Promise (namely, Save One) is really a supererogatory option, which shouldn’t be a duty even conditionally. This suggests a tweak:
Okay or Next Best*: PE(B/¬A) iff (i) PE(B), (ii) B is the best option compatible with ¬A, or (iii) the only alternative to B (that is compatible with ¬A) is supererogatory.
The tweaked principle gets the right answer in Kamm’s case: Do Nothing is permissible if you don’t Keep Your Promise, because the only alternative—Save One—is supererogatory. (That is to say, it’s non-conditionally supererogatory.)
But the tweak wreaks havoc in other cases. Sometimes, it gives us too many conditional permissions. Suppose you have three options: Keep Your Promise (permissible), Save One at great cost (supererogatory), or go on a Murderous Rampage (very wrong). Murder is impermissible, even conditional on not keeping the promise. But Okay or Next Best* implies that the rampage is conditionally permissible, since the only alternative left (Save One) is supererogatory. That is ridiculous.
In other cases, both versions of Okay or Next Best give us too few conditional permissions. They do this in cases of contrary-to-duty conditional supererogation.
Suppose Alice is safe and Betty is in mortal danger.Footnote 19 There are two buttons before you: Buttons A and B. Pressing either button will seriously harm Alice without her consent, causing her to lose her legs. But if you Press B, that will have two more effects: you will also lose your legs, and Betty’s life will be saved. While it’s true that if you Press B you will save Betty’s life, you are not required to do so. And pressing either button is wrong, given the harm caused to Alice. Now here is the key point. It is not true that, if you are going to harm Alice, then you must save Betty. You are not required to Press B conditional on pressing a button. Why not? Because sacrificing your legs to save a stranger is paradigmatically supererogatory and that is, effectively, what you are doing when you Press B rather than Press A. That is why we think Press B is not obligatory conditional on pressing a button. Instead, Press B is conditionally supererogatory. And since the condition is a wrong act (namely, button-pressing), we have an instance of contrary-to-duty conditional supererogation.Footnote 20 Even if one does not share these judgments, they are substantive judgments that are worth taking seriously; they should not be ruled out from the start by an account of conditional permissions and obligations.
This spells trouble for existing views of conditional obligation. Okay or Next Best implies that, if you are going to press a button, you must Press B, since that is the best option besides not pressing anything. We get the same result, for the same reason, from Okay or Next Best*. This seems extreme—a kind of fanatical moral offsetting. If you eschew the moral minimum, you may be obliged to make sacrifices that seem wildly disproportionate. (In Kamm’s case, a broken lunch date compels you to sacrifice your legs. Better not skip dinner!)Footnote 21 Again, this view is substantive and controversial; it shouldn’t trivially follow from our theory of conditional moral judgments.
There is an objection lurking.Footnote 22 Supererogatory acts are supposed to be good, but pressing button B is bad—how could a bad act be supererogatory, even conditionally? Our answer is that, on the standard definition, supererogation is not always good in some absolute sense; it is just comparatively better than a permissible alternative (see Muñoz, 2021b). Conditionally supererogatory acts, therefore, just need to be better than a conditionally permissible alternative, and Press B is indeed better than Press A.
In the All or Nothing Problem, it is wrong to save one stranger rather than both, even though saving one at great cost to yourself is better than saving zero. Likewise, in our Two Buttons case, pressing B (causing Alice to lose her legs, causing you to lose your legs, and saving Betty’s life) is wrong even though it costs you greatly and is better than pressing A (simply causing Alice to lose her legs). Both actions are unambiguously wrong. Each is decisively ruled out by an alternative. And yet, each represents a remarkably good sacrifice in comparison to a third option. That is the kind of case where we find conditional supererogation. Judgments of conditional supererogation guide agents toward the best remaining options while still acknowledging that betterness may come at a serious cost, which might justify refraining. So far, we have not found any principle that can make sense of wrong acts that are conditionally supererogatory.
5 A solution: justifying and requiring
Let’s take stock.
We have raised problems for two principles of conditional obligation. The first, Nothing Else Left, holds that we are conditionally obligated to pick from our remaining permissible options—if there are any. This principle fails in a case of consistent-with-duty conditional supererogation: the All or Nothing Problem, where saving one stranger is permissible, indeed supererogatory, conditional on not saving two. This problem can be solved with a second principle, Okay or Next Best, which conditionally permits saving one because it is the best option left, if you won’t save two. But this principle, even when modified, cannot handle contrary-to-duty conditional supererogation. Even if you wrongly do harm, you are not obligated to harm in the best possible way if the costs to you are disproportionate.
Why don’t these principles work? Nothing Else Left ignores betterness. In the All or Nothing Problem, if you will not save both, Nothing Else Left forbids Save One even though this is better than Do Nothing.Footnote 23 Okay or Next Best, meanwhile, ignores costs. In the Two Buttons case, if you are going to press a button, Okay or Next Best obligates you to Press B, even though the benefits of doing so (saving Betty) are not enough to outweigh the costs to you (losing your legs). If we are to make sense of these cases, we need a principle that is more flexible and powerful—something that directly factors in not only the justification we have to choose better options, but also the justification afforded by costs to the agent, whether those costs are suffered beyond the call (as in normal supererogation) or beneath it (as in contrary-to-duty conditional supererogation).
Thankfully, we do not need to start from scratch. The concepts we need are already there in the literature on reasons and supererogation. In particular, we will borrow the distinction between requiring reasons and justifying reasons. (Sometimes this distinction it put in terms of a reason’s “requiring strength” versus its “justifying strength.”) A requiring reason tends to make actions obligatory. A justifying reason merely tends to make actions permissible.Footnote 24 To illustrate, suppose I can save Chico’s life at the cost of my legs. I have a requiring reason to help Chico. If helping were costless, I would have to do it. But I don’t actually have to help, since that would cost me my legs, and I have a powerful reason not to harm myself. This reason isn’t itself a requiring reason. (More accurate: I have more justifying reason not to self-harm than I do requiring reason.) But my reason can still counterbalance the reason to help Chico, blocking a requirement to give aid.
The upshot of all this is that an act is permissible iff the justifying reason in favor can outweigh the requiring reason to do otherwise.Footnote 25 More officially:
J&R: PE(B) iff for any alternative A, the requiring reason to do A (rather than B) does not outweigh the justifying reason to do B (rather than A).
“J&R” is short for (you guessed it) “justifying and requiring.” We will assume that justifying and requiring reasons are somewhat, though not entirely, independent of each other. Any requiring reason doubles as a justifying reason, but not vice versa. There is always at least as much justifying reason to do an option as there is requiring reason. However, it is possible for the justifying reasons in favor of an option to outstrip the requiring reasons.Footnote 26
This is exactly what happens with supererogation. We already saw this in my simple choice between keeping my legs or saving Chico. But the real payoff will be applying J&R to many-option cases. So consider our first supererogation case: you can Save Zero (no cost), Save One (no cost), or Save Two (at the cost of your legs). Here, it is permissible and best to save both strangers. This is what there is the most requiring reason to do. It is also permissible to save just one, it seems, because you have a weighty justifying reason to keep your legs. But it would be wrong to save no one. There is more requiring reason to Save One instead, and no justifying reason to compensate. Save Zero and Save One are both worse than Save Two, but only the latter is justifiably worse, in the sense that it is worse but still supported by enough justifying reason to keep it permissible. What makes an option wrong is being unjustifiably worse than an alternative, i.e., not being supported by enough justifying reason to make up for the deficit in requiring reason.
With J&R in place, we can offer our principle of conditional permissibility.
Conditional J&R: PE(B/¬A) iff for any alternative C that is compatible with ¬A, the requiring reason to do C (rather than B) does not outweigh the justifying reason to do B (rather than C).
To see if B is permissible conditional on ¬A, we need to know how B compares to the options that are still being considered. If B is the best remaining option, or tied for best, it is permissible. If B is worse than some C, we have to ask: is it unjustifiably worse? If so, then B is conditionally wrong. If B is justifiably worse, then it may still be conditionally permissible.
The core idea here is that we have a two-step process for determining permissibility conditional on ¬A. First, remove A from the set of options. Second, take each option that remains, and ask whether it can be justified over each of the remaining alternatives; iff the answer is “yes,” the option is conditionally permissible. The conditionally permissible options are the ones that are permissible to choose from the restricted set of options. What does it mean to say that we “remove” the option of doing A, when we condition on ¬A? One possibility is that we imagine a counterfactual scenario in which the restricted menu is the agent’s entire option set, that is, we consider the case in which A is not available as an option at all. This is not how we see it. Rather than considering a counterfactual scenario in which A is off the menu, we are holding fixed A’s presence on the menu, but ignoring it in that we are not counting options as impermissible simply because they are unjustifiably worse than A.Footnote 27 In this sense we exclude the option of doing A from consideration, although we still suppose that A is available: we are still talking about a scenario in which the agent has the ability to do A.Footnote 28
With Conditional J&R laid out, we can next ask whether it gets the right answer in our hard cases. We believe it does.
First, unlike Nothing Else Left, our principle solves the All or Nothing Problem. Recall your three options: Save Zero (permissible), Save One (wrong), and Save Two (supererogatory). Conditional J&R has the plausible implication that saving just one is permissible conditional on not saving both, since saving one is the best option compatible with not saving both. Moreover, saving zero is also permissible conditional on not saving both, since it is justifiably worse than saving one (which is the only other option left). Conditional J&R succeeds where Nothing Else Left falls short.
Second, Conditional J&R does better than Okay or Next Best in Kamm’s case. Here your options are: Keep Your Promise (permissible), break your promise to Do Nothing (wrong), and break your promise to heroically Save One (supererogatory). According to Okay or Next Best, if you do not Keep Your Promise, you are obligated to Save One. This seems awfully demanding. Conditional J&R issues no such demand. For staying home is justifiably worse than life-saving (despite being unjustifiably worse than promise-keeping).Footnote 29
Finally, Conditional J&R is the first principle that can handle our Two Buttons case, where the options are: Do Nothing (permissible); Press A, harming Alice (wrong); and Press B, harming Alice and saving Betty at a big cost to you (also wrong). Conditional on pressing a button, it is intuitively optional, not obligatory, to Press B. Pressing B would be contrary-to-duty conditionally supererogatory. Conditional J&R has a neat explanation. Although it is worse to Press A than to Press B, pressing A is justifiably worse, given the cost to you of pressing B. Press B is thus conditionally supererogatory. It is conditionally permissible and better than a conditionally permissible alternative. Since the condition is a wrong act (namely, pressing a button), we have an instance of contrary-to-duty conditional supererogation.
This completes our argument for Conditional J&R. The view gives powerful explanations and plausible verdicts on cases beyond and beneath the call of duty. The view is also principled. It is not an ad hoc concoction, but the natural product of a view of conditionals and a view of supererogation: conditionals restrict quantification, and supererogation emerges from the clash between justifying and requiring reasons. We do not claim that Conditional J&R is the only principle to avoid embarrassment in the cases we have considered (one could mimic the results of Conditional J&R without appealing to the distinction between justifying and requiring reasons, and without even appealing to reasons at all).Footnote 30 But, in order to keep things tidy, Conditional J&R is the only such principle we consider here. There may be even lovelier principles left to discover. But given the simplicity and popularity of J&R, we think Conditional J&R is a natural place to start looking for a theory of conditional supererogation.
6 Conclusion
In this paper, we presented a series of puzzles for any theory of moral conditionals that ventures beyond obligation into the realm of the supererogatory. We began by presenting cases featuring consistent-with-duty conditional obligation (Sect. 2), consistent-with-duty conditional supererogation (Sect. 3), and contrary-to-duty conditional supererogation (Sect. 4), arguing that no existing principle can capture plausible intuitions about at least three of our cases. We then presented Conditional J&R and showed how it can capture plausible intuitions about all the cases (Sect. 5). Moreover, Conditional J&R has a principled rationale; it combines a familiar conception of conditionals (as restrictors on quantification) with the supererogationist’s insight that permissibility depends on the balance of requiring reasons and justifying reasons. It would seem that our examples—our three “hard cases,” and the more familiar cases of conditional obligation like the gentle murder paradox—for all their differences, can be given a surprisingly systematic treatment. Still, we do not claim that Conditional J&R is the only hope for understanding moral conditionals in the realm of supererogation. Our conclusion is more modest. We have found one way to solve the puzzles, though there may be more solutions, and indeed more puzzles, yet to be discovered. There may even be puzzles for our own view, and we may need to make revisions—but with any luck, they will be gentle.
Notes
We focus on conditionals whose antecedents, intuitively, restrict the options that an agent will consider in practical deliberation (e.g., “if you are not going to save both strangers, you must save one”), and whose consequents ascribe a deontic status to an option in their option set. We set aside “obligations,” stated with anankastic conditionals, that merely express how to achieve a certain end (e.g., “if you want to get the job, you have to interview”); see Greenspan (1975), von Fintel and Iatridou (2005), and Condoravdi and Lauer (2016). There is also a thriving literature on indicative conditionals whose antecedents provide information about the situation (e.g., “if the miners are trapped in Shaft A…”), and whose consequents involve obligations that are sensitive to that information; see Kolodny and MacFarlane (2010), von Fintel (2012, pp. 22–30), Cariani et al. (2013) and Charlow (2013). Another kind of conditional obligation, less central in moral theory, arises in what Schwager (2006, p. 242) calls a “relevance conditional,” whose antecedent “filters out one of the conditions (typically relevance) under which the speech act arising from an utterance of the consequent in the given context would be appropriate.” For example, “If I may be honest, you shouldn’t quit your day job.”.
Our work owes a debt to Paul McNamara, who has done more than anyone to develop a deontic logic for supererogation; see Mares and McNamara (1997), McNamara (1996, 2011a, 2011b). But his semantics in these papers, with its transitive ranking on the worlds permissible to bring about, is too restrictive for the puzzles in §§2–4. For his latest thinking, see McNamara (forthcoming).
We focus on “moral supererogation,” which is morally better than the moral minimum, as opposed to “rational supererogation,” which is rationally better than the rational minimum (see Benn & Bales 2019).
Horton (2017, p. 96). He states his principle in terms of contrastive obligation (to do one act rather than another), but elsewhere (p. 94) implies that it is equivalent to our formulation. (Note that he does not present Nothing Else Left as a general account of conditional obligations.)
When we say “permissible” (rather than “conditionally permissible”) outside of the consequent of a conditional, we mean plain old non-conditional permissibility. Mutatis mutandis for “supererogatory,” “obligatory,” and “wrong.” We prefer “non-conditional” to “unconditional” because “unconditionally permissible” sounds like it means “permissible no matter what,” which is not the intended meaning.
Horton doesn’t see this case as a counterexample to Nothing Else Left; he just uses Nothing Else Left to generate the All or Nothing Problem.
For early discussions of cases like this one, see Fried and Parfit (1979), Parfit (1982) and Kagan (1989, p. 16). For more recent discussions, see Wessels (2015), Pummer (2016), McMahan (2018), Sinclair (2018), Frowe (2019), Portmore (2019, §6.4), Bader (2019), Pummer (2019), Muñoz (2021c) and Rulli (2020).
Some writers argue that saving zero is morally better than saving only one, which suggests that if you won’t save two, you must indeed save zero. This is not our preferred solution to the All or Nothing Problem (for that see Pummer 2019; Muñoz 2021c), but it is an important contender, and it should not be ruled out a priori. Later we will show how to express this kind of view—defended by Lazar and Barry (ms.)—within our framework (fn. 30).
We use the dyadic conditional operator PE(_/_) without defining it in terms of a conditional and the monadic operator PE(_). We do not claim that such a definition is impossible; we just wish to abstract away from irrelevant controversies about, e.g., the proper syntax of ‘if’ (on which, see Kratzer 2012, Chapter 4). For simplicity, we talk as though the arguments of PE(_/_) are options, i.e., things that an agent could do, like murder gently. Officially, the arguments are propositions, like <Jack will murder gently>.
We assume that obligation and permission are interdefinable: what you are obligated to do is to pick some or other permissible option, and what you are permitted to do is to pick an option consistent with your obligations. PE(A) iff ¬OB(¬A). This rules out dilemmas, but dilemmas are not our topic here (for discussion of contrary-to-duty scenarios and deontic dilemmas, see Kiesewetter 2018). Note also that on a dyadic interpretation of conditional obligations, the inference from ‘OB(A or B)’ to ‘OB(A/¬B)’ is invalid (Pummer 2019, p. 286). This inference, as well as various forms of deontic detachment, is valid on wide-scope interpretations—‘OB(if not B, then A)’ is equivalent to ‘OB(A or B)’, where ‘if’ is read as the material conditional.
‘PE(B)’, of course, means that B is non-conditionally permissible. If desired, we could define this as permission conditional on a tautology: PE(B) = df PE(B/A or ¬A).
Note that this conditional permission (to save one stranger) is “contrary to duty” in one sense, but not in the sense that we have been using so far. The condition is not a wrong action; it is the permissible omission of a supererogatory action, namely, saving two. That said, the act of saving the one is itself wrong, non-conditionally. So the act is itself contrary-to-duty, even though the moral permission to do it is not conditioned on a transgression of duty.
See Pummer (2019, p. 285). As a reminder, here we are just assuming that Save Zero is worse than Save One. Not everyone agrees with this see the discussion of Lazar and Barry in fn. 8, above.).
The same implication follows from Nothing Else Left. If you do not keep the promise, there is only one permissible option left—saving the life.
Kamm writes: “The view is also problematic if it implies (and perhaps it does not) that if I fail to do my particular duty in order to do a supererogatory act, I am then obliged to do that supererogatory act. Although the pressure on me may increase to do the supererogatory act, I do not believe that failing my lunch date will necessarily leave me with a duty to give up my kidney to save someone” (1996, p. 318). (In her example, saving the life involves a kidney transplant rather than a rescue mission.).
Some will say that, since you must avoid the wrong of doing nothing, you are obligated to Save One if you do not Keep Your Promise. We disagree, but again, our framework is flexible enough to accommodate this intuition (see fns. 8 and 30).
Kamm’s case is famous because it seems to involve a nontransitivity: you may do nothing rather than save the life (in a pairwise choice), you may save the life rather than keep your promise, but you may not do nothing rather than keep your promise. For discussion, see Archer (2016), Portmore (2003, pp. 314–6, 2017), and Muñoz (2021c).
Even if it were obligatory to keep the promise in Kamm’s case, Okay or Next Best would still wrongly imply that you must save the life if you won’t keep the promise. Doing nothing isn’t “okay,” nor is it the best option after promise-keeping is excluded. For a similar case to this interpretation of Kamm’s case, see the “Two Buttons” example below.
This case is much cleaner thanks to comments from Kerah Gordon-Solmon.
In our Two Buttons example, the “contrary-to-duty” options all violate a deontological restriction—Alice’s right against harm. This is not essential to the case. In another version, suppose that you can either do nothing, Press C to costlessly save 100 lives, or Press D to donate your kidney to save a stranger named Debbie. You aren’t obligated to Press D if you won’t Press C. (Deontological restrictions are also inessential, by the way, to Kamm’s case, in which Do Nothing and Save One both break a promise. In another version, suppose that you are choosing between doing nothing, giving your legs to save a life, and costlessly saving another stranger’s finger from being crushed. The only permissible options are Save One and Save the Finger, but conditional on not saving the finger, Save One is supererogatory, not obligatory.)
There is also a worry about diachronic inconsistency. Consider the Two Buttons case. On the “offsetting” view, you have to sacrifice your legs to save Betty’s life if you are going to harm Alice. But if the choice to harm and the choice to save take place at different times, the choice looks radically different. Harming Alice on Monday doesn’t obligate you to heroically sacrifice for Betty on Tuesday. (This is especially clear if the harm to Alice is low and the cost of saving Betty is high.).
We owe this helpful objection to a anonymous referee.
There is also a more formal diagnosis of the trouble that Nothing Else Left has in the All or Nothing Problem. The case, as described, violates a principle that Sen (2017, Chapter 1.6*) calls Beta. According to Beta, if A and B are both permissible, then adding more options cannot make only one of them wrong. But this is what seems to happen in the All or Nothing Problem: adding the permissible Save Two makes Save One wrong while leaving Do Nothing permissible. In such a case, Nothing Else Left won’t work. The only permissible options are Save Two and Do Nothing, and yet you needn’t Do Nothing conditional on ¬Save Two. Nothing Else Left has trouble with Kamm’s case for the same reason: adding the permissible Keep Your Promise makes Do Nothing wrong while leaving Save One permissible.
By “tends to make,” we mean “contributes towards making,” not “usually makes.” Our distinction is based on work by Gert (2004, 2007); for a related distinction, see Greenspan (2005); for a moral version, see Portmore (2011). See also Hurka and Shubert (2012) on prima facie duties (which tend to favor and require) versus prima facie permissions (which tend to justify), as well as Muñoz (2021c) on moral reasons versus prerogatives. We could take (merely) justifying reasons as a primitive (Hurka & Shubert 2012), or we could try to understand them in other terms. One idea is that they act as “disabling conditions” on requiring strength (see Dancy 2004 on disablers). For a more developed view, see Muñoz (2021a) on moral defense.
We are using ‘reason’ here as a mass noun rather than a count noun.
Since this is a paper about conditional obligation, we do not aim to defend J&R at length over rival views of non-conditional obligation. But we should discuss one issue. We believe that an action’s choiceworthiness—how strongly “favored” it is—depends only on how much requiring reason there is to do it. Now, some will object that purely justifying reasons are also favorers. But if that is so, it is hard to see how supererogation could be favored over the moral minimum. The justifying reason to keep one’s kidney, if it can outweigh the requiring reason to donate, will also strongly favor selfishness (Hurka & Shubert 2012, p. 9). Another objection to our view is that some reasons are purely commendatory—they favor without justifying or requiring (Horgan & Timmons 2010; Archer 2016; Little & Macamara 2017). But this seems to invite odd recombinations. Suppose we sweeten a wrong act with commendatory reasons until it is the best option—even better than the permissible alternatives. Then the optimal option will be wrong. This seems absurd. A virtue of our view is that it rules out such possibilities.
Here is an example to illustrate why it might be important to hold fixed the availability of an excluded option. Suppose you have three options: dress up and attend a costume party (Costume), go to the party without dressing up (Casual), or skip the party and stay at home (Skip). Since Costume is an option, it would be wrong to pick Casual; it expresses disrespect to the host (let’s suppose). Intuitively, it is true here that you must not attend if you are not going to wear a Costume, i.e., if Costume is excluded. But if Costume were unavailable, there would be nothing disrespectful about Casual. The fact that one could dress up colors the choice between dressing casually and skipping out. That is why it is important not to imagine that Costume is unavailable when excluding it from consideration. (Our thanks to Joe Horton for this example and for many other helpful ideas.)
Because Do Nothing and Keep Your Promise are (roughly) equally costly, but much less costly than giving your legs to Save One, there is an interesting result. You have a powerful justifying reason to Do Nothing rather than Save One, but not to Do Nothing rather than Keep Your Promise. This is what Muñoz (2021c) calls a “comparative prerogative,” or what we might call a “contrastive justifying reason” (see Snedegar 2017 on contrastive reasons, but note that Snedegar’s (2021) treatment of supererogation does not commit to these reasons, or to any treatment of three-option cases like Kamm’s and Horton’s). Admittedly, “comparative prerogative” sounds like some heavy-duty jargon. But the idea behind it should be uncontroversial, and the term “comparative” is inessential. The point is just that relative costs are what determine justifying reasons (at least, the cost-based ones). Pointing to the costs of doing A, even if they are steep, will not justify doing B in the slightest, if B’s costs are just as big or bigger. (We would like to thank an anonymous referee, Justin Snedegar, Brendan de Kennessey, and Benjamin Kiesewetter for helpful comments here.)
Some people think there is always a (decisive) requiring reason to avoid doing what is non-conditionally wrong (Darwall 2010). We are skeptical of this claim, but we want to stress that it can still be combined with Conditional J&R. The result is a view that entails the principle we started with: Nothing Else Left. If only one permissible option remains, there is decisive reason not to pick anything else. In the case behind the All or Nothing Problem, for example, the view says you must Save Zero if you will not Save Two, because of the decisive reason against Save One. See Tucker (ms.) for a related view, on which there is a decisive requiring reason against doing wrong, but no obligation to Save Zero if you will not Save Two.
References
Archer, A. (2016). Moral obligation, self-interest, and the transitivity problem. Utilitas, 25, 355–382. https://doi.org/10.1017/S095382081600009
Bader, R. (2019). Agent-relative prerogatives and suboptimal beneficence. Oxford Studies in Normative Ethics, 9, 223–250.
Benn, C., & Bales, A. (2020). Rationally supererogatory. Mind, 129, 917–938. https://doi.org/10.1093/mind/fzz055.
Bonevac, D. (1998). Against conditional obligation. Noûs, 32, 37–53.
Cariani, F., Kaufmann, M., & Kaufmann, S. (2013). Deliberative modality under epistemic uncertainty. Linguistics and Philosophy, 36, 225–259. https://doi.org/10.1007/s10988-013-9134-4
Charlow, N. (2013). What we know and what to do. Synthese, 190, 2291–2323. https://doi.org/10.1007/s11229-011-9974-9.
Chisholm, R. M. (1963). Contrary-to-duty imperatives and deontic logic. Analysis, 24, 33–36.
Comesaña, J. (2015). Normative requirements and contrary-to-duty obligations. Journal of Philosophy, 112(11), 600–626.
Condoravdi, C., & Lauer, S. (2016). Anankastic conditionals are just conditionals. Semantics & Pragmatics, 9, 1–69. https://doi.org/10.3765/sp.9.8
Dancy, J. (2004). Ethics without principles. Oxford University Press.
Darwall, S. (2010). But it would be wrong. Social Philosophy and Policy, 27, 135–157.
Forrester, J. W. (1984). Gentle murder, or the adverbial Samaritan. Journal of Philosophy, 81, 193–196.
Fried, C., & Parfit, D. (1979). Correspondence. Philosophy and Public Affairs, 8, 393–397.
Frowe, H. (2019). If you’ll be my bodyguard: Agreements to save and the duty to minimize harm. Ethics, 129, 204–229.
Gert, J. (2004). Brute rationality. Cambridge University Press.
Gert, J. (2007). Normative strength and the balance of reasons. Philosophical Review, 116(4), 533–562. https://doi.org/10.1215/00318108-2007-013
Greenspan, P. S. (1975). Conditional oughts and hypothetical imperatives. Journal of Philosophy, 72(10), 259–276.
Greenspan, P. S. (2005). Asymmetrical practical reasons. In M. E. Reicher & J. C. Marek (Eds.), Experience and analysis (pp. 387–394). Öbv & Hpt.
Hansson, B. (1969). An analysis of some deontic logics. Noûs, 3, 373–398.
Horgan, T., & Timmons, M. (2010). Untying a knot from the inside out: Reflections on the ‘paradox’ of supererogation. Social Philosophy and Policy, 27, 29–63.
Horton, J. (2017). The all or nothing problem. Journal of Philosophy, 114, 94–104.
Hurka, T., & Shubert, E. (2012). Permissions to do less than best: A moving band. Oxford Studies in Normative Ethics, 2, 1–27.
Jackson, F. (1985). On the semantics and logic of obligation. Mind, XCIV(374), 177–195.
Kagan, S. (1989). The limits of morality. Clarendon Press.
Kamm, F. (1985). Supererogation and obligation. Journal of Philosophy, 82, 118–138.
Kamm, F. (1996). Morality, mortality: Volume II. Oxford University Press.
Kiesewetter, B. (2018). Contrary-to-duty scenarios, deontic dilemmas, and transmission principles. Ethics, 129, 98–115.
Kolodny, N., & MacFarlane, J. (2010). Ifs and oughts. Journal of Philosophy, 107(3), 115–143.
Kratzer, A. (1986). Conditionals. In A. M. Farley, P. Farley, & K. E. McCollough (Eds.), Papers from the parasession on pragmatics and grammatical theory (pp. 115–135). Chicago Linguistics Society.
Kratzer, A. (2012). Modals and conditionals: New and revised perspectives. New York: Oxford University Press.
Lazar, S., & Barry, C. (ms). Supererogation and optimisation.
Lewis, D. (1973). Counterfactuals. Harvard University Press.
Little, M., & Macamara, C. (2017). For better or worse: Commendatory reasons and latitude. Oxford Studies in Normative Ethics, 7, 138–160.
Mares, E., & McNamara, P. (1997). Supererogation in deontic logic: Metatheory for DWE and some close neighbours. Studia Logica, 59, 397–415.
McMahan, J. (2018). Doing good and doing the best. In P. Woodruff (Ed.), The ethics of giving (pp. 78–102). Oxford University Press.
McNamara, P. (1996). Must I Do What I Ought? (or Will the Least I Can Do Do?). In M. Brown & J. Carmo (Eds.), Deontic logic, agency and normative systems (pp. 154–173). Springer.
McNamara, P. (2010). A bit more on Chisholm’s paradox. Supplement to McNamara 2019.
McNamara, P. (2011a). Praise blame obligation and DWE: Toward a framework for classical supererogation and kin. Journal of Applied Logic, 9(2), 153–170. https://doi.org/10.1016/j.jal.2009.09.007.
McNamara, P. (2011b). Supererogation, inside and out: Toward an adequate scheme for common sense morality. In M. Timmons (Ed.), Oxford studies in normative ethics (Vol. 1, pp. 202–235). Oxford: Oxford University Press.
McNamara, P. (2019). Deontic logic. In E. N. Zalta (Ed.), The stanford encyclopedia of philosophy (Summer 2019 Ed.). https://plato.stanford.edu/archives/sum2019/entries/logic-deontic/
McNamara, P. (forthcoming). A natural conditionalization of the DWE-framework. In M. Brown, A. Jones, & P. McNamara (Eds.), Festschrift for Risto Hilpinen. Springer.
Muñoz, D. (2021a). From rights to prerogatives. Philosophy and Phenomenological Research, 102(3), 608–623.
Muñoz, D. (2021b). Infinite options, intransitive value, and supererogation. Philosophical Studies, 178(6), 2063–2075.
Muñoz, D. (2021c). Three paradoxes of supererogation. Noûs, 55, 699–716. https://doi.org/10.1111/nous.12326.
Parfit, D. (1982). Future generations: Further problems. Philosophy and Public Affairs, 11, 113–172.
Portmore, D. (2003). Position-relative consequentialism, agent-centered options, and supererogation. Ethics, 113, 303–332.
Portmore, D. (2011). Commonsense consequentialism: Wherein morality meets rationality. Oxford University Press.
Portmore, D. (2017). Transitivity, moral latitude, and supererogation. Utilitas, 29, 286–298. https://doi.org/10.1017/S0953820816000364
Portmore, D. (2019). Opting for the best: Oughts and options. Oxford University Press.
Pummer, T. (2016). Whether and where to give. Philosophy and Public Affairs, 44, 77–95.
Pummer, T. (2019). All or nothing, but if not all, next best or nothing. Journal of Philosophy, 116, 278–291.
Rulli, T. (2020). Conditional obligations. Social Theory and Practice, 46, 365–390.
Schwager, M. (2006). Conditionalized imperatives. In C. Tancredi, M. Kanazawa, I. Imani, & K. Kusumoto (Eds.), Proceedings of SALT XVI (pp. 241–258). CLC Publications.
Sen, A. (2017). Collective welfare and social choice: Expanded Edition. Harvard University Press.
Smith, M. (1993). Violation of norms. In Proceedings of the fourth international conference on artificial intelligence and law (pp. 60–65). Amsterdam, The Netherlands.
Snedegar, J. (2017). Contrastive reasons. Oxford University Press.
Snedegar, J. (2021). Reasons, competition, and latitude. In R. Shafer-Landau (Ed.), Oxford studies in metaethics (Vol. 6, pp. 134–156). Oxford University Press.
Sinclair, T. (2018). Are we conditionally obligated to be effective altruists? Philosophy and Public Affairs, 46, 36–59.
Tucker, C. (ms). The weight of reasons. (Tentative title.)
von Fintel, K. (2012). The best we can (expect to) get? Challenges to the classic semantics for deontic modals. Central APA.
von Fintel, K., & Iatridou, S. (2005). What to Do If You Want to Go to Harlem: Anankastic Conditionals and Related Matters. MIT.
Wessels, U. (2015). Beyond the call of duty: The structure of a moral region. Royal Institute of Philosophy Supplement, 77, 87–104.
Acknowledgements
We would like to thank two anonymous referees for their thoughtful and constructive comments. We also thank Ralf Bader, Joe Bowen, Roger Crisp, Brendan de Kenessey, Kerah Gordon-Solmon, Peter Graham, Brian Hedden, Joe Horton, Benjamin Kiesewetter, Paul McNamara, Per Milam, Tina Rulli, Justin Snedegar, Jack Spencer, Chris Tucker, and our audience at the All or Nothing Workshop Miniseries. (Special thanks to those who realized that, even if they weren’t going to every talk in the series, they were still allowed to go to ours!)
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Muñoz, D., Pummer, T. Supererogation and conditional obligation. Philos Stud 179, 1429–1443 (2022). https://doi.org/10.1007/s11098-021-01724-y
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11098-021-01724-y