Does morality demand our very best? On moral prescriptions and the line of duty

Abstract

It is widely accepted that morality does not demand that we do our very best, but our most significant moral traditions do not easily accommodate this intuition. I will argue that the underlying problem is not specific to any particular tradition. Rather, it will be difficult for any moral theory to account for binary moral concepts like permissible/impermissible while also accounting for scalar moral concepts like better/worse. If only the best is considered permissible, morality will seem either unreasonably demanding or implausibly minimal. But if we draw a line of duty below the optimal, then we must explain how the act that is worse is nonetheless permissible. Some have tried to explain this by appealing to non-moral considerations, and others have appealed to agent-relative moral considerations. I argue that no such approach will work. We should instead exploit the distinction between reasons for performing an act and reasons for holding someone accountable for an act’s performance. This approach will also help to clear up a confusion regarding the notion of a moral demand.

1 Introduction

There are many examples of acts that seem to go beyond the demands of morality, acts that philosophers call ‘supererogatory.’ Some of these involve taking extraordinary risks or making profound sacrifices. Consider the case of Wes Autrey who jumped onto the New York City Subway tracks in order to save a man who had fallen in front of an oncoming train. Acts like these are exceptional of course, but they are hardly unheard of. Moreover, there are a great many less dramatic examples of people doing good beyond what duty requires, examples like shoveling a neighbor’s walk or cooking a meal for a friend who’s going through a tough time.

Of course, proclaiming one’s own conduct beyond the call of duty is something of a social faux pas (‘I really did not have to do that, how extraordinary!’). And so it is not uncommon for the agent herself to deny in these cases that she has done anything more than was her duty. Wes Autrey, for example, said: “I don’t feel like I did something spectacular; I just saw someone who needed help. I did what I felt was right (Buckley 2007).” But these protestations are usually thought the result of humility overwhelming honesty; we do not take them at face value. The belief remains that these acts go beyond duty, that one can do more than morality demands.

But despite the strength of this intuition, it is not easily accounted for within our major moral theories. Others have noted the difficulties of squaring supererogation with the particular claims made, for example, by Kantians or utilitarians.¹ I will argue that the challenges faced by these particular theories ride atop a deeper problem for moral theorists, a challenge that any moral theory will need to confront. This is the challenge of accounting for both the binary and the scalar elements of our moral discourse.² That is, it is difficult for a moral theory to account plausibly for either/or concepts like permissible and impermissible on the one hand while also accounting for degreed concepts like better and worse on the other. But supererogation—where the supererogatory act is better than the minimally permissible act—involves both of these sets of concepts. The difficulty this raises for moral theory is clear if we look at the Kantian and utilitarian traditions. Kant gives us an account of duty but no clear grounds for assessing an act as better than what duty requires. Utilitarians give us an account of acts as better and worse but no clear grounds for saying an act is permissible though less than the best.

The utilitarian assesses acts by reference to the goodness of their consequences, and a standard utilitarian account of right action claims that an act is right if and only if its (expected) consequences are at least as good as those of its alternatives. But then nothing less than the best will do, and if we draw the line of duty at the point of optimal goodness, what results is obviously an extremely demanding morality. Indeed, it may be so demanding as to make our binary moral concepts of little actual use. It is certainly plausible to think that almost none of our actual acts were the very best available. Or at least it seems we could have little confidence that they were. If so, an optimizing account of permissibility will make the distinction between permissible and impermissible of little use in our moral assessments.

In fact, Alastair Norcross has argued that utilitarians should simply claim that there is no sharp distinction between right and wrong. They should, he claims, embrace the notion that rightness, like goodness, is a scalar concept, that acts can be more and less right and that there is no line at which an act goes from right to wrong. For whatever line we might try to draw, the utilitarian will be just as concerned about increases in goodness above the line or below the line as he will about increases in goodness from below to above the line. As a result, Norcross argues that utilitarians should “reject the claim that duties or obligations constitute any part of fundamental morality (Norcross 2006, p. 43).”³ He writes that “consequentialist theories such as utilitarianism are best understood purely as theories of the comparative value of alternative actions, not as theories of right and wrong that demand, forbid, or permit the performance of certain actions (2006, p. 38).” And so the utilitarian does not offer an account of permissible and impermissible to go along with his account of better and worse, and so the utilitarian will not be able to account for supererogation.⁴

Kant’s view has, in a way, the opposite shortcoming. Kant provides an account of a line of duty but no grounds for judging acts as better and worse above that line. Of course, Kant claims that an act has moral worth only if done from duty, and so it would seem that there can be no morally good action beyond duty. But even if we were to set this claim aside, Kant’s theory simply provides no grounds from which we might assess an act as better than duty. Kant’s moral law places a constraint, or a limiting condition, on our pursuit of our aims. The Categorical Imperative offers a test for an action’s permissibility, and our duty is determined by reference to this moral law. But an act will either pass or fail the test the Categorical Imperative provides—the maxim is universalizable, or it isn’t; the act involves treating someone merely as a means, or it does not. Of course we may say that the impermissible are generally worse than the permissible, but this will not capture the sense in which goodness is a matter of degree.

Thomas Hill has suggested that Kant might account for the idea that we can do better than duty requires by appealing to the concept of an imperfect duty and to the concept of a wide imperfect duty, in particular (1992, pp. 147–175). Imperfect duties, in general, allow more latitude in how they are carried out than do perfect duties. For instance, while basic respect for others may always be required, there will usually be a great many ways to satisfy this requirement. Wide imperfect duties allow more latitude not just in how they are carried out but in when they are carried out as well. For instance, the requirement that one adopt the maxim of beneficence is consistent with our frequently omitting possible beneficent acts as long as we perform enough such acts on other occasions. Kant denies, however, that once we have done enough to satisfy the imperfect duty, we could still do better. But if so, then adopting the maxim of beneficence will be either unreasonably demanding or implausibly minimalist. That is, either we accept that doing enough will involve doing all the good it seems we could do, or we deny that it would actually be better to go beyond some reasonable minimum.

There is good reason to think that Kant endorsed this minimalist position. While Kant certainly did not believe that we do our best by promoting the best consequences, he may well have shared with the utilitarian the intuition that we have a duty to do our very best. In the Doctrine of Virtue, for example, he claims that “the first command of all duties to oneself” is to “know yourself… in terms of your moral perfection (1797, 1964, p. 441).”⁵ We are under a duty to perfect ourselves as moral agents. But duty doesn’t involve performing so-called super-meritorious acts only because Kant denies that they are actually better. For Kant, it is not okay to do less than the best, but the best involves less than one might think.

Kant was well aware of the common notion that there are acts that go beyond duty, but he was not impressed by it. In the Critique of Practical Reason, he writes, “I wish they [educators] would spare them [students] examples of so-called noble (super-meritorious) actions… and would refer everything to duty only… (1788, 1993, pp. 160–161)” Kant is here making a point about moral instruction. He claims that examples of the super-meritorious lead students to flights of moral fancy and self-righteousness and that such students will be prone to “release themselves from observing common and everyday responsibilities as petty and insignificant (1788, 1993, p. 161).”⁶ But Kant’s concern is not merely for good moral pedagogy. Of the drive to be especially noble, Kant claims that even “among the instructed and experienced portion of mankind, this supposed drive has, if not an injurious, at least no genuine moral, effect on the heart… (1788, 1993, p. 163)”. We should “attend not so much to the elevation of the soul… as to the subjection of the heart to duty” (1788, 1993, p. 161).

So Kant’s theory provides an account of permissible and impermissible but no obvious account of better and worse. Utilitarianism, on the other hand, provides an account of better and worse but may give us no useful account of permissible and impermissible. This is not to suggest that no account of better and worse is possible in a generally Kantian framework or that no account of permissible and impermissible is possible in a generally utilitarian framework. But providing such accounts is not easy. As we’ve just seen, if we claim that only the best is permissible, then ‘permissible’ and ‘impermissible’ will be of little use in distinguishing between our various actions since it is quite possible that they are (almost) all impermissible. But if we draw the line of duty below the optimal, then we face the challenge of explaining how the act that is worse is nonetheless permissible since it appears we have good (sometimes quite weighty) reason to prefer the act that is better. That is, doing what is worse appears to involve a moral (and rational) failing. And it may well be this underlying problem that drives both Kant and the utilitarian to deny that one can exceed morality’s demands. But put this way, the problem is not unique to utilitarianism or Kantianism. It is a structural problem that faces any attempt to account for the notion that we can do better than morality demands.

2 Some potential solutions

Though I do not have the space to consider many of the proposed solutions to this problem,⁷ I will briefly discuss a few models that may be thought to resolve the problem and that are, I think, representative of certain broad approaches that one might take. In doing so, I hope to motivate the idea that a very different kind of solution may be required.

2.1 Ethical contextualism

As noted above, Alastair Norcross claims that binary moral concepts are not part, as he puts it, of “fundamental morality.” But while maintaining that these concepts are not a part of fundamental moral theory, Norcross has elsewhere argued that we can nonetheless make room for these concepts in our ordinary moral discourse (Norcross 2005a).⁸ To do this, he argues for a position he calls “ethical contextualism.” On this view, terms such as ‘rightness’ and ‘goodness’ are not taken to refer to properties of the act itself, but they are instead defined in such a way that the truth-value of our judgments of rightness and goodness will vary according to the linguistic context within which they are issued. He defines ‘right,’ in particular, in the following way: “An action is right iff it is at least as good as the appropriate alternative (Norcross 2005a, p. 89).” But what counts as the appropriate alternative will depend on the conversational context.

To clarify, Norcross uses an example he calls Burning Building. In the example, we are to imagine a case in which there are ten people trapped in a burning building and an agent who can rescue them; but the agent can only rescue one at a time, and each individual rescue involves considerable effort and risk. Norcross notes that Burning Building looks like a good example of a case in which duty does not require maximizing. He writes that in order “to get a context that would set the rescue of all ten people as the appropriate ideal, we could imagine a conversation among committed maximizing consequentialists, or perhaps among proponents of a Christ-as-ideal moral theory, or perhaps it will be enough to imagine a conversation in a philosophy class that has just been presented with maximizing consequentialism (2005a, p. 89).” He adds that a classroom context involving the discussion of the demandingness objection to consequentialism “might set a pretty lax standard.” On this view, the truth-value of the statement “the agent’s act was (at least) minimally permissible (right)” will depend on the appropriate alternative set by the given linguistic context.

A difficulty with this approach is that many of the practices that attach to our binary moral judgments—assigning punishment and blame, taking on guilt, or paying compensation—will inevitably extend across conversational contexts.⁹ Indeed, I will argue in Sect. 3 that our judgments of moral permissibility are deeply connected to our practice of holding ourselves and others accountable. And while it seems plausible to say that our judgment of an act as right or wrong may depend on the linguistic context, it is harder to accept that whether or not I should, say, take on a burden of guilt varies with the linguistic context. And so to adequately capture our binary moral judgments within a contextualist account, we will at least need to allow for a privileged context within which we ought to make those judgments that will issue in particular moral sanctions. And so we’re still left with the question of what determines that privileged context and how it is that the appropriate alternative is less than the best. In general, for any given context in which the appropriate alternative (the minimally permissible act) is not the optimal act, we can still ask what factors make it such that the appropriate alternative is less than the ideal. We can still ask how it is that it is permissible to do worse when one could do better. But then contextualism alone will not fully answer the challenge of accounting for the notion of the minimally permissible.

2.2 Appeal to the non-moral reasons against

Susan Wolf considers the idea of a moral perfectionist, or a moral saint, and she argues that “such a being would be unattractive (1982, p. 419).” Given the great many opportunities that we have in life to display moral virtues, Wolf fears that any serious attempt at moral perfectionism would require an agent to develop moral virtues such that they would crowd out non-moral virtues and that it would lead an agent to spend so much time performing moral acts that there would be little or no time left for engaging in other worthwhile endeavors. As a result, a moral perfectionist would be unlikely, according to Wolf, to have a “healthy, well-rounded, richly developed character (1982, p. 421).” She claims that the desire to be as morally good as possible could be all consuming in a way that would threaten “the existence of an identifiable, personal self.”¹⁰ Wolf is not arguing against any moral theory but rather against the idea that it is always most appropriate to fully satisfy the demands of morality (whatever they may be). Her point is not that we have the wrong moral ideals but rather that we need to recognize that they are not our only ideals and that it is both desirable and appropriate to give a significant regard to non-moral ideals as well.

Though Wolf suggests that accepting her conclusion means accepting that “any plausible moral theory must make use of some conception of supererogation (1982, p. 438),” there is reason to think that supererogation may be a separate issue altogether. Wolf claims that sometimes non-moral considerations make it appropriate to not act on the weight of moral reasons, but she does not claim that such a failure to act on the weight of moral reasons is appropriate because it is morally permissible. She doesn’t argue that it is morally permissible to not be as morally good as one can be but rather that it is not “particularly rational or good or desirable” for one to be that morally good (1982, p. 419). This does not provide a framework for an account of supererogation. If she is right, then there is a sense of the word ‘ought’ that attaches to the normativity of reasons generally (including non-moral reasons) such that there are occasions on which what one ought to do, generally speaking, is not what one ought to do, morally speaking. This is certainly an interesting claim. However, we are still left with the question of whether it is ever morally permissible to not do what it would be morally best to do.

Even if we accept that non-moral reasons can change the moral status of an act—that they can make an act morally permissible despite the moral reasons that tell against it, we are left with serious problems.¹¹ First, it is not clear that we could, on this model, properly limit the scope of the permissibly sub-optimal. While it does seem in some cases that the costs of the morally optimal act make it permissible to omit the act, the costs do not seem to have that effect in other cases. Consider the costs associated with not stealing things. Even if I would really like a sweater, it still does not seem permissible for me to swipe it.¹² Thus if we allow non-moral reasons to override moral reasons, we would need to explain why they can override some but not others. Second, the appeal to non-moral reasons will only be plausible in cases where there are strong non-moral reasons to omit the morally better act. But it is just not the case that all (or even most) supererogatory acts involve significant costs to the agent. Consider the example of buying a book for a friend. Suppose it’s on sale, and you think your friend will enjoy it. It may be morally better to buy it than not, but it is the nature of such favors that they are not obligatory. Such acts do not involve significant sacrifice and may even be in the interest of the acting agent if the joys of giving outweigh the costs. But if the non-moral reasons and the moral reasons both favor giving (or even if the non-moral reasons are just not strong enough to override), then on the view we are considering, the agent would be morally required to do so.

2.3 Appeal to the moral reasons against

Jonathan Dancy argues that in addition to its neutral value—“the value derived from the difference it will make to the world that this action was done (1993, p. 138)”—an act also has an agent-relative value that is generated by the sacrifice incurred (or presumably the benefits gained) by the acting agent. Of course, the costs to the agent, as one person that will be affected by the act, will already be accounted for in the act’s neutral value, but Dancy claims that the costs to the agent can, in some sense, be counted twice. The cost to the agent contributes to the moral value of the act considered neutrally, and it also contributes to the moral value of the act when considered from the perspective of the acting agent.¹³

But how does this relate to the problem of accounting for supererogation? Consider the example of a soldier who might jump on a grenade in order to save two others. That the soldier will lose his life will contribute negatively to the neutral value that the act has. However, this negative contribution will (other things equal) be outweighed by the positive contribution resulting from the fact that two other soldiers will survive when they otherwise would not have. Thus, agent-neutral reasons will likely favor the act. But Dancy’s picture requires us to consider the agent’s loss, not simply as a loss in the calculation of neutral values, but also as a loss to the agent. That is, we must also, according to Dancy, consider the agent-relative reasons involved, and it may turn out that the agent’s reasons, all told, do not recommend the act, although his reasons rooted in the neutral value of the act do recommend it. Thus, it is at least plausible that the agent will have more moral reason to run for cover than to jump on the grenade. If this is the case, then running for cover will certainly be a permissible option. However, as Dancy himself is well aware, this way of looking at the situation may go too far in that it now appears as though the soldier ought to run for cover (1993, p. 140).

Dancy claims that this problem can be dealt with because, although the costs to the agent lead to moral reasons, the agent can simply discount the costs to himself since it is he that will have to pay them. This is so despite the fact that those costs are moral costs and that agent-relative values are moral values. As Dancy notes, it is important to his account that the costs involved generate a moral reason so that agent-relative costs give rise to a genuine permission and not simply an “excuse for the faint-hearted (1993, p. 213).” On the model that Dancy gives us then, the soldier has more moral reason to run for cover but may discount the costs to self such that jumping on the grenade is also morally permissible.

However, there remain several shortcomings in Dancy’s account. First, as we have already seen, many acts that are taken to be supererogatory simply do not involve significant costs to the agent even while they do involve significant neutral value. Thus in a great many cases of seemingly supererogatory acts, it is hard to believe that the costs to the agent, even if valued in two ways, could significantly impact the value of the act, all told. Indeed, some acts that are regularly considered supererogatory are actually beneficial to the agent. Consider again the case of buying a book for a friend. Such a favor may be morally optimific and also be quite pleasant for the agent, and yet it would hardly seem obligatory. But valuing the costs to the agent twice in these cases will simply add to the reasons in favor of the supererogatory act. For an example in which the costs to the agent exist but seem insignificant, consider the case of contributing to UNICEF instead of buying new tennis shoes. Such charitable contributions (at least beyond a certain minimum) are widely regarded as paradigm examples of the supererogatory, but Dancy’s theory does not account for this intuition. Assume that the money, if donated to UNICEF, will be well spent on potentially life-saving efforts in impoverished communities and also assume the agent already has a pair of reasonably acceptable, if not particularly great, tennis shoes. If the money is donated, the agent will forgo the value created by the improvement in his foot apparel, and on Dancy’s model, this cost must be considered both neutrally and as a cost to the agent. But for Dancy’s model to work in this case, we would have to accept that this cost, when valued both ways, would give the agent more moral reason to buy the tennis shoes than he has to make the donation, despite the significant agent-neutral moral reasons generated by the potentially life-saving benefits of the donation. This seems unlikely. Consider also the stock example of supererogatory heroism: the soldier jumping on the grenade. If doing so would save two others, then it seems quite plausible to say that the cost to self, if valued twice, would give the soldier more moral reason to dive for cover than to smother the grenade. But what if he could save ten, or twenty others? Does the act then become obligatory, or can we say that his life, when valued both ways, outweighs the agent neutral cost of those many other deaths? This seems implausible. It also hints at the second problem with Dancy’s model.

The second problem with Dancy’s model is that even those cases that would plausibly count as supererogatory on his view will count in a sort of backwards way. In the UNICEF example, we saw that Dancy can allow that both giving to UNICEF and buying the shoes are permissible only if buying the shoes is preferable when all values are counted fully. Even if it were the case that the benefits of buying the shoes, when valued twice, make buying the shoes preferable, this would seem to make giving to UNICEF merely permissible and not supererogatory. The supererogatory act, then, would be buying the shoes since it is the act that is better (than the minimally permissible). Dancy’s model threatens to elevate acts that seem merely permissible to the level of the supererogatory and to make seemingly supererogatory acts merely permissible. He could object that the supererogatory act should simply be identified with the act that would be required if agent-relative costs were discounted. But given that agent-relative costs generate moral reasons on Dancy’s view, this would imply that when all the moral reasons are on the table, a supererogatory act would necessarily be morally suboptimal—it could not be the morally preferable course.

2.4 Appeal to the reasons for

One response to the challenge that supererogation poses might be to simply reject the optimizing conception of reasons that underlies the problem; one might endorse a satisficing conception instead. In first introducing the satisficing position, Michael Slote uses the example of an agent selling a house (Slote 1989, p. 9). He claims that it may be perfectly reasonable to decide at the outset what would count as an acceptable offer (depending on “what he has paid for the house, what houses cost in the place he is relocating, and on what houses like his normally sell at”) and to then accept the first offer that meets the established criteria. One need not seek the best possible offer. Thus Slote argues for a satisficing, rather than an optimizing, account of reasons. He argues that it is perfectly reasonable to choose an inferior, though satisfactory, option.

The challenge posed by supererogation is to explain how the supererogatory omission is permissible, how it is permissible to take the lesser course and not the course that is favored by the weight of moral reasons. And so in answer to this challenge, the defender of satisficing might argue that an agent, when faced with a supererogatory option, need not act on the weight of reasons because that is just not how reasons, in general, work.

But consider again Slote’s example. Given the costs of keeping a house on the market indefinitely, even the committed optimizer may fully endorse Slote’s approach to selling a house on the grounds that a satisficing approach is the best optimizing strategy.¹⁴ If you can never be sure that the offer in hand is the best you will get, at some point, you have to just take an offer that is good enough. On the other hand, if you have two offers side by side, one better than the other, it sure seems you ought to take the better offer. But then where satisficing makes sense in this case, it does not involve choosing an admittedly inferior option but rather it amounts to doing the best one can with limited information.¹⁵ It is easy to see how it is permissible to use a satisficing strategy in this way; it is much harder to see how it could be perfectly rational to take the worse option when the better option is right before you.

But rather than saying that rational agents, in general, need not act on their best reasons (but need only do well enough), perhaps we should instead distinguish between two kinds of reasons, some of which can generate requirements and some of which cannot. That is, perhaps some reasons will fail to generate a requirement not because they are defeated but because they are just not the kinds of reasons that lead to requirements. If so, we may be able to account for supererogation if the reasons that recommend the supererogatory act are of this non-requiring sort.

Jonathan Dancy, for example, has argued for a distinction between peremptory, or requiring, reasons on the one hand and enticing reasons on the other. While it may be irrational to not act on certain reasons (like a reason to get your brakes checked), it is not irrational to “do a less enticing act when there is a more enticing one available (2004, p. 91).” Enticing reasons “make the action they recommend worth doing, fun, exciting, attractive, and so on (Dancy 2004, p. 99).” But they are not the kind of reasons that generate requirements, and so enticing reasons can be permissibly ignored. Dancy himself does not believe that the notion of an enticing reason is useful in explaining moral supererogation since a moral reason is, he believes, by nature a requiring reason.¹⁶ It is the nature of a moral reason that it would generate a duty if it were the only relevant consideration.¹⁷ Even acts that seem typically supererogatory cannot permissibly be omitted where this would be wholly gratuitous. But Joshua Gert has proposed an account that avoids this difficulty.

Dancy claims that a reason may be an enticer in some contexts but not others, but he also claims that when it is an enticer it is not, in the same context, a requirer. Rather than viewing reasons as either requirers or enticers, Gert suggests that we should view them as having two dimensions of normative strength: requiring strength and justifying strength.¹⁸ The requiring strength of a reason to ϕ allows it to play a role in making it irrational to not-ϕ. The justifying strength of a reason to ϕ allows it to play a role in justifying one’s ϕing where it would otherwise be irrational.

Gert gives the following useful example. I am rationally required to not put two hundred dollars in a box that is an incinerator; the cost makes it irrational for me to do so. If the box is an alms box instead, then the benefits to the poor will justify the costs and so make the act rational. But it does not follow that the act is required, just that it is justified. In fact the reasons in support of giving could justify a great deal. Say the two hundred dollars could prevent serious malnutrition in a number of children; that is the kind of consideration that could plausibly justify risking one’s life (if there is no easier way to bring about the result). On the other hand, the reasons against the donation involve the loss of two hundred dollars. But one is certainly not justified in risking one’s life to save two hundred dollars. So the reasons for the donation have more justifying strength than the reasons against, but the “insufficiency of this altruistic reason to rationally require the donation is the result of its not having a great deal of requiring strength (Gert 2003, p. 9).”

Gert claims that altruistic reasons, in particular, have more justifying than requiring strength. It may be reasonable not to act on them even where they have greater justifying strength than the reasons in support of the alternatives because they do not have a lot of requiring strength. But he does not deny that they have requiring strength. In fact, he admits that “perhaps all reasons have some requiring strength (Gert 2003, p. 17). Certainly if one could provide great benefits to others at no or entirely trivial cost to oneself, then one ought to do so. But Gert’s claim is that the costs to oneself can justify not doing so because they have enough justifying strength to overcome the requiring strength of the altruistic reasons. It is the justifying strength of the opposing reasons that matters. So the alternative act may be supported by even less requiring reason than the altruistic act. If one is considering buying a pair of shoes or donating to childhood hunger relief, it certainly seems there would be little or no requiring strength to the reasons supporting the shoe purchase (it doesn’t take much at all to justify doing otherwise). And the altruistic reasons will have greater justifying strength as well. As Gert says, one may even be justified in risking her life to prevent children from starving, and surely such risks are not warranted in order to get oneself a pair of shoes. But then the reasons in support of the altruistic act have greater requiring strength and greater justifying strength. The reasons for giving are better all around.

Nonetheless, Gert denies that there is a rational failing involved in not giving. It does not matter that the donation is, in a way, more justified. What matters is that the justifying strength of the reasons to do otherwise is sufficient to overcome the requiring strength of the reasons to give. If the justifying strength on each side is strong enough to outweigh the requiring strength on the other, then both options are justified and neither is required. The trouble with this answer is that it suggests that reason is neutral between two acts once the requiring reasons have been overcome. It suggests that reasons act like on/off switches. If an act has enough requiring reason, that switches other acts to irrational. If an act has enough justifying reason, it then switches from irrational to rational. These, on Gert’s view, are the two roles that reasons play. But the challenge supererogation poses arises precisely because reasons do not have just a binary function but a scalar function also. They do not merely justify (or not) or require (or not) acts; they recommend or favor them.¹⁹ And where the difference in rational support is substantial, the recommendation may be strong. But then, why is it not a rational failing when one ignores the stronger reason, when one fails to do as reason (including the requiring and justifying parts) strongly recommends? Of course Gert will say that it is not the case that you ought to do as reason recommends if there is good enough justifying reason to do otherwise. But the question we are concerned with is how this reason (despite being less strong) can be good enough to play that role, and Gert’s account does not answer it.

3 Beyond the balance of reasons

I have argued that moral theorists face a serious challenge in accounting for the strong intuition that we can do better than morality demands—the intuition that morality does not demand our very best. That is, they face a serious challenge when they attempt to account for the supererogatory. If the supererogatory act is better than the minimally permissible act, then we need to explain how it is that the act that is worse is nonetheless permissible. It seems we will have good moral reason to prefer the act that is better; so how is it not a moral failing to do what is worse instead? And if our challenge is to explain how it is permissible to omit the better (the supererogatory) act, it makes perfect sense that we would try to answer this challenge by looking at the reasons that count against it. But the problem is that any such approach will get the default wrong. It is not the case that the supererogatory omission requires a justification such that we need some good reason to omit the supererogatory act, but rather, part of what is interesting about the supererogatory act is that one often does not need any good reason to omit it. Shoveling a neighbor’s walk (assuming that the neighbor is perfectly able to do this himself) or buying a gift for a friend is not obligatory regardless of whether I have any great reason not to. But then any appeal to the reasons against the supererogatory act will not really explain why its omission is permissible. But we also cannot simply look at the reasons for the act and claim that they are not the sort that generate requirements. The reasons in support of a supererogatory act seem, very often at least, to be of the sort that can lead to a requirement. And so we are left with the question of why they do not generate an obligation or a rational requirement in cases where the act is supererogatory.

To adequately account for the supererogatory, I believe we need a very different approach.

Instead of looking at the reasons for and against the supererogatory act, we need to appeal to a distinction between two sets of reasons: the reasons for and against the act on the one hand, and the reasons for and against holding the agent accountable for the act on the other. This is a distinction that has been noted by a number of utilitarians, including Sidgwick and most notably Mill,²⁰ but the importance of the distinction for moral theory (utilitarian or otherwise) has not been widely recognized.

In chapter five of Utilitarianism, Mill suggests that an act cannot be a moral duty unless some punishment, or punitive sanction, should be attached to its omission. He writes that “we call any conduct wrong, or employ, instead, some other term of dislike or disparagement, according as we think that the person ought, or ought not, to be punished for it; and we say that it would be right to do so and so, or merely that it would be desirable or laudable, according as we would wish to see the person whom it concerns, compelled, or only persuaded and extorted, to act in that manner (1861, 2006b, p. 246).” This suggests that Mill is allowing for the category of supererogation as distinct from moral duty. Mill writes this passage while attempting to mark off our common idea of justice from more general moral categories so that he may address the question of how our idea of justice relates to utility. And given that Mill claims to be endorsing a utilitarian conception of morality, with which a clean distinction between morality and expediency is difficult to square, it would be tempting to read Mill here as simply discussing a common feature of our moral talk and not as endorsing supererogation as a distinct moral category. However, in “Auguste Comte and Positivism,” written shortly after Utilitarianism, Mill makes clear his endorsement of some kind of supererogation as a distinct moral category:

It is not good that persons should be bound, by other people’s opinion, to do everything that they would deserve praise for doing. There is a standard of altruism to which all should be required to come up, and a degree beyond it which is not obligatory, but meritorious… The object should be to stimulate services to humanity by their natural rewards; not to render the pursuit of our own good in any other manner impossible, by visiting it with the reproaches of other and of our own conscience (1865, 2006a, p. 338).

Beyond this, unfortunately, Mill says little about the concept of supererogation and its relation to utilitarianism. But what he says here is quite suggestive. Note that Mill is focused not principally on what the agent ought to do but rather on what the agent ought to be bound (by the opinions of others and the reproaches of conscience) to do. Thus we can read Mill as arguing that there are some acts such that an agent ought to perform them and yet no punitive sanctions should apply if the agent fails to perform them. And the punitive sanctions that are prohibited (that it would be inappropriate for the agent or others to introduce in order to promote the performance of the meritorious act) go well beyond those encompassed by our practices of blame. The prohibited sanctions would include anything that would stand as a motive by causing the agent to be averse to the prospects of omitting the act, rather than as a direct, positive motive to perform the act, and they would include internal punitive sanctions like guilt and remorse. Whether or not this is the best way to read Mill, I will argue that it offers the beginnings of a very promising account of supererogation.

The distinction between the reasons for action and the reasons for holding people accountable for those actions connects with another distinction between a prescriptive moral ought and a moral obligation. This latter distinction is frequently invoked in our common moral discourse—consider: ‘you really ought to visit your mother on her birthday, but you have got to at least send a card.’²¹ But it has been largely ignored by moral theorists. That is, while moral theorists often treat prescriptive oughts and obligations as identical, our common sense morality recognizes an important difference between them. The difference is that an obligation is conceptually tied to accountability and to holdings accountable, to what you have got to do; that is, you are not obligated to perform an act unless you can properly be obliged to perform it, or be held accountable for its performance (whether by demands or normative expectations, or by punitive sanctions in the case of failure, sanctions like punishment, blame, or even guilt). But it is not the case that we can properly be held accountable for everything that we ought morally to do. And this is because an agent’s reasons for performing an act are not the same as our (or even the agent’s²²) reasons for holding the agent accountable for the act’s performance.

With this in mind, we can begin to answer the challenge as it was initially posed—the challenge of plausibly accounting for acts as better and worse while also plausibly accounting for a line of duty. An act is better or worse by virtue of the reasons for and against it, and in the end, we should accept that we ought to do what we have the most reason to do—we ought to do our best. But the line of duty is determined not by whether we ought to perform the act but by whether we ought to be held accountable for its performance. There will be cases then in which one’s obligations involve doing somewhat less than she could do even if she really ought to do her very best.

The reasons we have to hold agents accountable for action may be tied to the risks of allowing certain behaviors to go unchecked. As T.M. Scanlon puts it, “we can see the need for limits on certain patterns of action… by seeing the ways in which we are at risk if people are left free to decide to act in these ways (1998, p. 201).”²³ Given the general interest in avoiding injury, we will have very good reason to enforce norms against violence, for example.²⁴ Given their importance in supporting valuable coordination between persons, we have reason to hold people accountable for keeping to their promises or following through on their commitments.²⁵ In general, we have good reason to enforce norms that protect against the callous disregard for the interests of others or that ensure the adequate consideration of others’ interests. These reasons may connect also with considerations of moral fairness. That we must demonstrate an adequate regard for the interests of others does not imply an individual obligation to optimally promote those interests. Rather, it involves recognizing a shared responsibility to promote those interests (Murphy 2000, p. 75). Each will ideally be required to do her fair share.

But given the limitless opportunities we have for promoting the interests of others, it seems that it would be morally undesirable to hold agents accountable for being optimally beneficent. If we were accountable for doing our very best, then few of our decisions would be protected from morality’s demands, and our freedom to pursue personal projects would be undermined. In fact some will argue that my own view does not go far enough in protecting that freedom. I will conclude with a brief response to that objection.

4 Is this still too demanding?

My view allows us to say that Kant is right to deny that duty involves doing all the good that we could do and that the utilitarian is right to assert that we ought to do our best. The mistake that they both make is to conflate the notion of duty with that of a prescriptive ought. Distinguishing between the two allows us to recognize the limits of duty without denying that we often have good reason to do better. One reason that some will be hesitant to endorse this approach is that it may, despite the limits placed on duty, seem to make morality, nonetheless, too demanding.

A central reason that we are concerned with supererogation is that it is important for allowing options in the face of admitted value (one may omit an act he rightly believes it would be better to perform). Some claim that to do this we must deny that we morally ought to do our very best; we must deny that there is any moral failing when we do not. They suggest that our freedom to pursue our personal projects requires protection from conclusive moral reasons to do otherwise.

But a prescriptive ought is simply not in conflict with options, and the belief that there is a conflict stems from a confusion with respect to the nature of moral demands. Talk of morality’s “demands” involves a useful metaphor, but it can be misleading. A moral prescription does not itself constitute a genuine demand. As we have seen, there may be conclusive reasons for an agent to ϕ even where it is not appropriate that anyone demand that she ϕ or that anyone demand an excuse (or sanction her) if she does not. Moral reasons sometimes license the issuing of moral demands—we will have good moral reason to hold agents accountable for certain acts. But moral reasons are not themselves demands. And while our freedom does require protection from demands, it does not require protection from moral reasons. Where one is not accountable for doing as she ought, she is perfectly free to choose otherwise. And this should not surprise since we probably do so with some regularity.