1 The Problem Stated

Typically, a political community does not simply implement rules, laws, and commands. If it claims authority, it also demands deference. And if an authority is legitimate, one ought to obey its directives, irrespective of the weight of some of the reasons not to do so (Raz, 2009a [1979]; Hart, 1990 [1982]; Simmons, 2016). Political authority therefore can be the site of moral conflict, where:

  1. 1.

    One (prima facie) ought to defer to authority: !(α);

  2. 2.

    One (prima facie) ought to act as the more weighty reason demands: !(β);

  3. 3.

    One can do either: ◊(α ˅ β);

  4. 4.

    One cannot do both: ¬◊(α ˄ β).

When faced with this dilemma, and irrespective of how it is resolved, can one be left with regret? Can it ever transpire that, whatever decision one makes, one will fail to do what one ought to do?

Some will give a negative answer to this question. Moral conflicts leave no remainder, no regret, they claim. Moral conflict sceptics of this kind begin from the ‘ought implies can’ principle, which requires that, if one ought to β it must be that one can β (Brink, 1994, pp. 226-7). Of course, a situation can arise where one (prima facie) ought to β, one (prima facie) ought to α, one can do either, and one cannot do both. Nonetheless, it is argued, what matters is the all-things-considered judgement of what one ought to do (Ross, 2002 [1930]; Conee, 1982; Sinnott-Armstrong, 1985; Foot, 2002 [1983]; 2002 [1995]; MacIntyre, 1988). If one (all-things-considered) ought to α (O(α)), and if one acts as one ought, the argument goes, there is no reason to feel regret. The other arm of the dilemma (!(β)) is not only defeated. It is no longer binding; it is cancelled. The defeated reason must be cancelled, it is claimed, so as to avoid the following logical inconsistency, where one cannot both β and α and yet one ought to both β and α (Donagan, 1993; de Haan, 2001; Horty 2003; McConnell, 2014).

What we have above is a general position of moral conflict scepticism. Many reach the same sceptical conclusion when considering authority, namely that there is no such dilemma here. One line of argument is that authority, and the deference it demands, can never have legitimacy in the first place (Wolff, 1998 [1970]; Gans, 1986; Moore, 1989; Regan, 1989; Alexander, 1990). If it is never the case that one ought to defer to authority, the dilemma cannot arise. But here I focus on those who accept it is possible for authority to have legitimacy, and yet who conclude that authority does not leave any remainder (Sect. 3 and Sect. 4). According to what I refer to as the reason-conforming approach, when deference to an authoritative directive is legitimate, acting in this way also better meets the demands of freedom and rationality (Raz, 1985, 1986, 2006). As a result, the requirement to act on the weight of reasons is cancelled, as follows:

  1. 1.

    One (prima facie) ought to defer to authority: !(α);

  2. 2.

    One (prima facie) ought to act as the more weighty reason demands: !(β);

  3. 3.

    One can do either: ◊(α ˅ β);

  4. 4.

    One cannot do both: ¬◊(α ˄ β);

  5. 5.

    One (all-things-considered) ought to defer to authority: O(α);

  6. 6.

    The requirement to act on the weight of reasons is no longer binding: ¬Binding(β).

In this paper, I try to show why it is that authority can leave a remainder. Before starting to advance that argument, however, it is worth noting that the more general position of moral conflict scepticism has been thrown into question already. Moral conflicts leave a remainder, it is argued, in those instances where a prima facie ought is defeated by the all-things-considered ought but the defeated reason is not cancelled. Then it is rational to feel regret when one fails to act as one (prima facie) ought (Sartre, 1957 [1946], pp. 295–7; Williams, 1965, p. 113; Marcus, 1980, p. 130; Gray, 2000, p. 9; Marino, 2015, pp. 29–31). The feeling of regret is rational because, first, the moral dilemma itself is not illogical. This is because the existence of separate oughts need not create a ‘conjunctive’ ought (Cf. Brink, 1994, p. 230). One ought to take each of two conflicting courses of action, and one can do either (Marcus, 1980, p. 134). But it does not follow one ought to do both. Hence, the charge of inconsistency can be avoided. Second, regret is rational in a moral dilemma if one has failed to act on a binding reason. To feel no regret is not an option, unless one is willing to be indifferent to the reason not acted upon, and when the reason in question is moral, this would involve moral indifference (Williams, 1965, p. 113). As a result, the dilemma is tragic, for no matter what one does, one does some wrong, including when one acts as one ought all-things-considered (Williams, 1981 [1976], p. 27; 2009 [1993], p. 208, n.10). And we see these conflicts not only in drama, in particular Greek tragedies (Nussbaum, 1986, p. 34), but also in real world political decision making, when, as the holder of a political office, a morally good person is forced to get their hands ‘dirty’ (Walzer, 1973, p. 167; Williams, 1978, p. 63).

As I said, my focus here is not moral dilemmas generally, but the specific issue of authority and the dilemmas to which it can give rise. And I want to consider cases where an authoritative directive has legitimacy, but it demands acting in ways that are morally objectionable. Then, whatever it is that one (all-things-considered) ought to do, one should feel regret for failing to act as one (prima facie) ought. And my argument here is that while such conflicts simply remain hidden from view for those who adopt the reason-conforming approach, this is not the case for another way to think about authority, what I call the norm-based approach, according to which legitimacy is a matter of adequate normative justification rather than better meeting the demands of freedom and rationality (Sect. 5). As some of its proponents acknowledge, the demands made by authority may be out-weighed by conflicting reasons, and yet the authority retains legitimacy (see Hart, 1955;, 2012 [1961]; Simmons, 1979). This is where we see an important difference between the two approaches. It is not that those who adopt a norm-based approach to authority must of necessity also reject moral conflict scepticism. Nonetheless, unlike the reason-conforming approach, a norm-based approach is at least compatible with an argument acknowledging the dilemma of authority.

2 The Dilemma of Authority Introduced

Whether or not authority can leave a remainder, in the sense discussed above, nonetheless, it is accepted that an authoritative directive can be in conflict with other reasons, and indeed with reasons of a different kind. Indeed, the requirement that one defer to a legitimate authoritative directive is explained by the distinctive way in which an authoritative directive operates as a reason for action.

An authoritative directive is, firstly, a content-independent reason. This is represented by the idea one should obey a command because it is a command, and independent of the nature of the action commanded (Hart, 1990 [1982], p. 101); and that one has a duty to obey the law simply because it is the law (Simmons, 1996, p. 24; Raz 2001, p. 9). A legitimate authoritative directive is also, secondly, an exclusionaryFootnote 1 reason. It is both a first-order reason for action (a reason to φ) and a reason not to act on some or all reasons to φ and to ¬φ (Raz, 2009 [1979], p. 17, p. 18; see also Hart, 1990 [1982], p. 101; Hart, 2012 [1961], p.81; Simmons, 2016, p. 13). A directive is different to a mere first-order reason, for example, the advice one receives. One should comply with advice insofar as it out-weighs other competing considerations. All that matters here is the merits of the case. An authoritative directive, in contrast, is a higher-level reason, a reason to exclude from consideration the weight of certain other reasons. And that explains how authority demands deference. As the focus here is on practical authorities, this is deference regarding how one acts, rather than what one believes. One is free to think what one likes about a directive, but one is to act as stipulated, and do so independent of the content of the action and regardless of the weight of some of the reasons not to act in this way. This stands in sharp contrast to those situations where one is left free to act on one’s own judgement of the merits of the case, as where one decides whether or not to accede to the advice proffered by a friend.

Authority demands deference. It does not follow that just anything can count as an authoritative directive. An authoritative directive is not a content-independent, exclusionary reason unless it has legitimacy. If one is subject to merely de facto authority, one may have good reasons to comply (sometimes moral, sometimes prudential) but not a duty to obey. In contrast, those with legitimate, de jure, authority have the right to issue laws (or rules, orders, and so on), and those subject to that authority have a duty to obey. Of course, it is true by definition that one has a legal obligation to comply with legal commands.Footnote 2 At issue here is whether one has in addition a moralFootnote 3 duty to obey (Hart, 2012 [1961], p. 8, p. 194ff). If one does, then one ought to defer to an authoritative directive.

However, even if we accept that authority can have legitimacy (that is, when the requirements of legitimacy are met), there is also reason to think that deference itself is morally objectionable (see Raz, 2006, p. 1014). Why is this the case? It is due to the conviction that freedom and rationality make the following demands: that one act on one’s own judgement; and that one act on the weight of reasons. The two combined require that one act as the more weighty reason demands. And the demands of freedom and rationality seem incompatible with authority, which not only entails deference, but can at times call on one to act in ways one knows to be mistaken.

In addressing that dilemma, it must be stressed, it arises regarding practical rather than theoretical authority. While practical authority relates to reasons for action, theoretical authority, in contrast, merely concerns reasons for belief (see Raz, 2009 [1979], p. 8; Simmons, 2016, p. 14; Peter, 2017; Christiano, 2020). For example, a chemist with expertise in the paint solvents used in the automobile industry is a theoretical authority on this subject, and so their authoritative utterances count as reasons for belief in this area. If the owner of an automobile factory receives advice from such an expert on the environmental damage that a particular paint solvent can cause, they should follow the advice insofar as the reasons offered outweigh other, competing reasons, including reasons derived from a purely economic consideration of the costs associated with any changes to the production process if the advice is followed. But that just means there is no deference to theoretical authority here: the factory owner is free to make their own decision, based on a consideration of the merits of the case. In contrast, practical authority is a reason to act in some way. For example, those with legitimate practical authority over legislation governing the use of chemicals in the workplace have the right to decide which chemicals can be used, by whom, and under what conditions. If they issue a rule forbidding the use of a certain paint solvent, the factory owner has a duty to obey, and to do so regardless of some of the reasons they have (including financial) not to do so.

When the dilemma of authority arises, therefore, it does so concerning practical authority. The first arm of the dilemma is the demand for deference, justified by the reasons explaining why the practical authority in question has legitimacy. The second arm of the dilemma is the requirement that one satisfy the demands of freedom and rationality, and so act on the weight of the reasons that apply. If authority has legitimacy, then deference is justified: one ought to defer. But how can deference be justified when, it seems, it conflicts with one’s freely acting on the basis of the more weighty reason(s)? That is our question. And there are, broadly speaking, three different approaches taken in response: (1) that the dilemma never arises, because authority never has legitimacy; (2) that the dilemma arises, but can be resolved without remainder; and (3) that the dilemma arises, but it can leave a remainder.

The first simply rejects legitimate authority, and in doing so removes the dilemma itself. Authority, the argument goes, is incompatible with freedom and rationality, and as a result, it can never be de jure: it is not the case that one ought to defer to authority. The first part of the argument is that those who issue rules do not offer reasons for action. They instead claim authority, and the claim resides in who they are, not in the content of their rules (Wolff, 1998 [1970], p. 6). Second, authority is incompatible with freedom. One is, as an autonomous agent, responsible for the rules one lives by; and if one obeys a rule independent of what one thinks about it, one is, the argument goes, reneging on those responsibilities as an autonomous agent (ibid., p. 12). For one version of the argument against authority, namely a priori philosophical anarchism, authority is unacceptable because it is incompatible with autonomous consent. Others reject authority without endorsing the anarchist sceptical conclusion that there are no legitimate political obligations now. For non-anarchist critics of authority, on occasion, a rule, say the requirement to pay one’s taxes, should be treated as if it were a higher level reason, but this is only because what it demands is of great significance, or weight. What is more, it ceases to be a rule (or a quasi-rule or ‘indicator rule’) when it conflicts with other more weighty reasons, here the reasons one has not to pay one’s taxes (see Gans, 1986, p. 388; Moore, 1989, p. 895; Regan, 1989, p. 1009; Alexander, 1990, p. 6). For the non-anarchists who reject authority, therefore, what matters is the weight of reasons to act as a rule demands. But that is why they are in agreement with a priori philosophical anarchists in contending that one should never defer to a rule.

This is an argument rejecting authority itself. In both versions of that argument, above, the dilemma of authority can never arise in the first place. But this argument appears to be falsified by our own experience. For what we often encounter is the phenomenon of the higher-level reason with very little weight. If the law demands that one stop at a red light, for example, one should defer to this rule. One should exclude other, conflicting reasons, such as one’s own judgement of whether traffic lights are the safest and most efficient way to deal with traffic. At the same time, the conflicting reasons can be weighty, as when one runs a red light rushing to bring an accident victim to hospital. Therefore, our experience suggests that rules operate by excluding reasons. But our experience also shows that any one rule has a certain weight, and at times it is defeated by a conflicting reason that happens to be more weighty (see Simmons, 1979, p. 25, p. 28).

That is why, for many, we cannot do away with the idea of authority, and the associated idea of legitimate deference. The implication is important. It is that there are two different ways in which reasons can be defeated: by being outweighed but also by being excluded. Then, it is possible the demand for deference will conflict with the requirement to act as the more weighty reason demands. One question this raises is, when the two conflict, what ought one to do all-things-considered? Here, my focus is a different one. I ask, can one be left with regret even when one does what one ought to do all-things-considered? If a defeated (prima facie) ought is still binding, it is not cancelled: one ought to act as it requires. But one cannot do so if one acts as required by an (all-things-considered) ought. Can such a dilemma arise regarding authority?

3 The Reason-Conforming Approach

The question is whether authority leaves a remainder, where it is rational to feel regret for failing to do what one ought to do. According to the reason-conforming approach, there is, in short, no such remainder. Deferring to a legitimate authoritative directive in fact better meets the demands of freedom and rationality. This is because legitimate authority better ensures conformity with reason, as any conflicting reason is not only defeated but also cancelled. There are various examples of this approach, including many of the different versions of deliberative democracy (see Manin, 1987; Christiano, 2004; Estlund, 2008). What they have in common is the assumption that what a legitimate authority demands is both what is rational and also what will guarantee freedom. I focus on a different version of that argument here: Joseph Raz’s ‘service conception of authority’. Authority ‘serves’ one, in the sense intended by Raz, not by promoting one’s interests, or satisfying one’s wants and desires. It serves one, rather, by helping one act on reasons that already apply (1986, p. 56). And that, for Raz, is what gives authority its legitimacy: legitimate authority helps one’s ‘rational capacity whose function is to secure conformity with reason’ (2006, p. 1017).

How does authority better ensure one conforms to reason? To answer that question, we need to consider Raz’s distinction between dependent and pre-emptive reasons. To begin with, ‘the justification of the binding force of authoritative directives rests on dependent reasons’, he says, which are reasons that ‘already independently apply’ to those who will be obligated (1986, p. 47, p. 59). One example he offers concerns the legal rules requiring parents perform their duties of care to their children. Here, the duties are reasons that already independently apply to a parent, and that is why they can be authoritative directives. However, if authoritative directives are based on some of the reasons that already independently apply, why should one not simply (directly) act on the dependent reasons? As a parent, should one not simply care for one’s children? Is there any need in addition for the directive? There is, the argument goes, because it is also pre-emptive: it has become the reason for action, and not the dependent reason(s) upon which it is based. The distinction is needed because an authoritative directive will be the outcome of a process whereby various and varied dependent reasons are balanced; and so, it will reflect one possible judgement on how best to balance some of these reasons (ibid., p. 46). For example, a parent’s duty of care will be informed by paternalistic commitments, along with considerations of the child’s right to make their own informed decisions, and also the challenge of reconciling present wishes and future needs, and so on (Fives 2017, pp. 201–2). This points to the variety of (dependent) reasons relevant to the duty of care. However, when a law demands that one care for one’s child in a particular way, say by accepting their right to make medical decisions without parental involvement, the intention is that one will act on the directive and not one’s own judgement of how one would balance those same reasons.

So, an authoritative directive is based on dependent reasons, and is itself a pre-emptive reason. That brings us closer to explaining why, for Raz, any conflict concerning an authoritative directive can be resolved without remainder. Because an authority is legitimate insofar as it ‘serves’ one in this way, one’s deference to an authoritative directive, he contends, will also be compatible with the demands of freedom and rationality. First, according to what Raz refers to as the ‘normal justification thesis’, one ‘would better conform to reasons which apply to [one] anyway (that is, to reasons other than the directives of the authority)’ if one intends to be guided by the authority’s directives than if one does not (2006, p. 1014; see 1986, p. 53). And second, according to what he calls the ‘independence condition’, ‘the matters regarding which the first condition is met are such that with respect to them it is better to conform to reason than to decide for oneself, unaided by authority’ (2006, p. 1014). Crucially, the two are not ‘incommensurable’, according to Raz (ibid., p. 1015). The concerns behind independence are ‘concerns with satisfying reasons’: and so, if an (alleged) authority does improve one’s conformity with reason, it will also be in accordance with why we value the ability to exercise our own judgement independently (ibid., p. 1016). And we value the ability to exercise our own judgement, Raz claims, ‘because of its purpose, which is, by its very nature, to secure conformity with reason’ (ibid., p. 1017).

Raz’s argument is that a legitimate authority better ensures one conforms to reason, and in that way satisfies the demands of freedom. Raz is interested in the ‘normative’ sense of reason: the reasons one should find binding, even if one fails to do so (2011, p. 18). Nonetheless, he does not go so far as to maintain that some of the reasons one happens to have are bad reasons: that they reflect nothing more than ideology and bias, for example. Other proponents of the reason-conforming approach do make that claim, and conclude that what matters is the reasons we arrive at after a transformative process of democratic deliberation (Manin, 1987, p. 364; Peter, 2021, p. 403). As I said, Raz does not place that kind of restriction on what can count as a reason. He simply assumes that each person has already-existing reasons for action, and it is these that are relevant when considering whether authority is justified. But, as we saw, he also assumes that the authoritative directive replaces some of one’s dependent reasons with a pre-emptive reason (1986, p. 60). It creates a higher-order (a second-order) reason for action. The authoritative directive stipulates what one is obliged to do, independent of the content of the action and regardless of the weight of some of the reasons to act or not to act in this way.

Raz’s ambition is to show that legitimate authority better ensures conformity with reason. This is a challenge, because even when an authority is legitimate, it might be that one has good (weighty) reasons not to obey its directives. This kind of conflict might arise where, say, a promise has made one duty bound, but one is then obliged to act in ways that are morally reprehensible, as when one has promised to obey a regime that then implements unjust laws (Simmons, 1979, pp. 78–79). But, as Raz clearly believes, such a conflict will not arise. This is because authority, Raz assumes, is not justified simply insofar as it possesses the requisite normative basis (here, respect for autonomous consent). It should be noted, this is a controversial claim. Most of those who have grappled with this question disagree, for they believe that their first task is to identify the normative grounds of legitimacy, whether consent, fairness, justice, membership, promoting the common good, and so on. Raz rejects that approach because, as we shall see, for him, any such normative basis is not sufficient for legitimacy. For example, a promise does not oblige, he insists, if the promised action does not better ensure one conforms to reason. But Raz is also saying an authority has legitimacy just insofar as its directives do better ensure one’s conformity with reason. Conformity with reason, therefore, is both necessary and sufficient for legitimacy.

Raz gives the example of ‘Colin’, ‘who promised his wife that in all decisions affecting the education of his son he will act only for his son’s interests and disregard all other reasons’ (2009 [1975], p. 39). This includes disregarding the reasons he [Colin] has to leave his job so as to devote himself instead to his love of writing. Perhaps Colin’s love of writing weighs heavily for him. But its importance is immaterial now. For Raz, Colin is duty bound, but it is not Colin’s promise that, by itself, explains why this is so. Indeed, a promise ‘is binding only if the promised action is of a class regarding which there are sufficient reasons to hold the promisor bound by the promise’. More generally, the promised act must belong to the class of actions that ‘enhances people’s control over their life to be able to make such promises’, and also it must not be ‘grossly immoral’: that is why ‘a promise to a slave is not binding, nor a promise to make someone else a slave’ (2006, p. 1013). Applying these points to the analysis of authority, the implication is that, one is bound to obey not simply or solely because of the normative consideration offered as a justification of authority (say, respect for autonomous consent), but because one’s deference better ensures one conforms to reasons that already independently apply, including the reason one has to retain one’s autonomy to act on reasons.

One implication of adopting the reason-conforming approach is that authority leaves no remainder. Raz claims to explain why one better satisfies the demands of freedom and rationality simply by obeying authoritative directives. His intention is to account for the reality and the legitimacy of authority and the deference it demands. It is important to stress this point. As we saw (Sect. 2), others agree with Raz that what matters is whether rules better ensure conformity with reason, but unlike Raz they conclude that authoritative directives cannot do so. Thus, they conclude, it is the weight of a reason, and it alone, that matters. Raz, in contrast, is arguing that rules can have legitimate authority; one should defer to them. But he is arguing that authority can have legitimacy in this way only if it better ensures one conforms to reason and in so doing meets the demands of freedom as well.

According to this line of thought, there is no genuine dilemma of authority, where no matter what decision one makes one is left with regret for what one ought (prima facie) to have done. We see this in one of Raz’s examples, where a regime imposes a heavy tax burden so as to help pay for the collective provision of social services (Raz, 1985, p. 148). What binds one, he says, is the reasons one has to do one’s share in providing this kind of service. Let us develop the example further. Imagine that, as a result of submitting to this tax law, one has less private income to use as one had planned, including the money that, on the advice of a financial expert, one wanted to put away for one’s own retirement. Raz would assume there is no dilemma here, and that one does nothing wrong when one obeys the rules of the regime, and as a result fails to save for retirement as one had good reasons to do.

However, there is a problem with Raz’s argument. He maintains that the normative qualities characteristic of a regime (for example, that it is based on respect for autonomous consent) are not sufficient to confer legitimacy. What matters, he argues, is whether one better conforms to reason by deferring to its directives. But insisting that conformity with reason is sufficient for legitimacy has the following paradoxical implication. It would entail that one is bound to obey any directive that better ensures one conforms with reason, including, for example, when a friend, unsolicited, issues rules on how best to save for one’s retirement. The fact that one’s friend’s rules are sound is not sufficient to show one is answerable to them, and so, one should obey them (Darwall, 2011, p. 104). More generally, for the directives of a regime to have authority, one must first have an obligation to act as the regime demands (Marmor, 2011, p. 127). And one explanation for why one would be obliged in this way is that the regime has certain normative qualities, as we shall explore in the final section. Therefore, it seems Raz is wrong to assume conformity with reason is a sufficient condition for legitimacy. But there is another serious problem with Raz’s position. It is his assumption that conformity with reason is also a necessary condition of legitimate authority. This is also a mistake, as I hope to show now.

4 Whither Deference?

The problem with the reason-conforming approach concerns the way in which it understands deference. More precisely, on this approach there appears to be no room for deference itself, or the dilemma to which it gives rise. This is clear when we consider what Raz has to say about what is arguably the quintessential instance of that phenomenon: the moral duty to obey the law. His argument is that there is a moral duty to obey the laws of a regime only insofar as doing so better ensures conformity with reason. And there are, Raz believes, only three scenarios where the law is likely to have these implications for reason, and hence have legitimate authority.

First, one has obligations to obey many of the regulations formulated by experts, Raz contends. The obligation is justified by their expertise, their ‘superior knowledge’, concerning the practice to be regulated, and where significant harm can be caused by not following these regulations, as is the case in a number of areas: safety at work, manufacturing standards, planning, professional qualifications, and so on (1985, p. 147, p. 148). Second, one has a moral duty to obey laws that solve coordination problems (1986, p. 56). Again the obligation arises from the requirement to ‘secure conformity with reason’, here because of what Raz believes is the law’s ability to achieve goals one has reason to pursue as an individual, but cannot do so without securing coordination from others. This is the case when, inter alia, public projects and welfare services are being funded (1985, p. 148; 2006, 1022). And a third justification is that one should obey the law when disobedience ‘will undermine the government’s ability to do good’ (1985, p. 146). As we all have reason to do what is good, the third justification of the duty to obey the law has the same rationale as the first two. What matters, again, is the reasons one has to do what the government demands.

However, although any of these could ground an individual’s duty to obey the law, Raz’s conclusion is that there is no general duty to obey the law. The argument from superior knowledge does not apply to anyone with greater expertise than those setting the rules; laws coordinating the efforts of large masses of people are inevitably simplified, and so some may find that they do not have reasons to follow certain aspects of those laws in a given situation; and finally, law breaking will not undermine the effectiveness of government in all cases, freeing people from the obligation to obey when that is the case (Raz, 2009 [1979], p. 239; 1985, p. 148, p. 149). Some obligations ‘probably’ exist and apply equally to all, Raz concedes, in particular the duty to pay one’s taxes (based on the coordination argument) and to refrain from political terrorism (based on the bad example argument) (1985, p. 149). But, beyond this common core, Raz concludes, obligations vary from person to person, and from occasion to occasion. And they vary because one is obligated only insofar as being bound in this way better ensures one’s conformity with reason.

What does this tell us about deference? I think this shows that Raz’s reason-conforming approach has taken deference out of the picture altogether. This is because he is saying one has obligations only where deference better ensures one’s conformity with reason. As Raz says, a legitimate authoritative directive will be ‘on the winning side of the argument’, as is the case with a tax to pay what one owes as part of the cost of community services (2006, p. 1022). The implication appears to be that one is not to act as required by an authority without first ensuring that its directives are based on reasons on the winning side of the argument. Of course, it is not his intention to remove deference in the way I am suggesting has in fact happened. It happens inadvertently. But when we examine his position, the upshot is that directives do not have authority. Rather than defer, one is to accede, and do so only regarding those directives that better ensure one’s conformity with reason.

One possible defence of Raz’s approach is to say that it explains when an authority has legitimacy in regard to certain areas of life, but not others. Where deference to the rules of authority does better ensure one conforms to reason, deference is justified. Within such a sphere, one is not to assess each and every rule, one by one, to determine if it is on the winning side of the argument. For, as Raz himself says, this would be contrary to the ‘whole point and purpose of authorities’ (1986, pp. 47–8). However, this line of defence does not hold up. In fact, Raz does not make a distinction between spheres where an authority has legitimacy and spheres where it does not. Instead, he leaves it up to each individual to discern for themselves when (or where) authority has legitimacy. But that just means there is no deference at all. As Raz himself maintains, there are some things about which an authoritative directive cannot be mistaken: ‘Mistakes which they [authorities] make about factors which determine the limits of their jurisdiction render their decisions void’ (ibid., p. 62). But who is in a position to judge whether the authority has made such a mistake? As we already saw, even though expertise is one of the bases of legitimate authority on Raz’s schema, one is not to defer to the rules set by experts when one has greater expertise on that topic, and one knows the directives are mistaken. That is, each one of us decides whether the authority exceeds its limits. The implication is that one does not defer to authority, even in those spheres where Raz believes it does (potentially) have legitimacy. One may or may not accede to its demand, and only after careful examination of the content of the demand.

Where does this leave the dilemma of authority? As we know, some maintain that moral conflicts are not real: they are not dilemmas that leave regret (Sect. 1). In a moral conflict, if one acts as one ought, the other arm of the dilemma is not only defeated but also cancelled. Raz’s work on authority arrives at the same conclusion, namely that the dilemma of authority does not leave regret. But his efforts do not succeed. He had hoped to show that an authoritative directive has legitimacy only insofar as deference better ensures conformity with reason. Conformity with reason is necessary and sufficient for legitimacy, he presumes. The problem is that this account leaves no place for deference itself. But is there an alternative? What is required is an approach that accepts the reality of deference. However, as we shall see, if we accept the reality of deference, then we are faced with the genuine possibility that, no matter what one does, one will be left with regret.

5 The Norm-Based Approach

I have already referred to a norm-based approach to authority. It starts with the question of how to justify legitimate authority, and hence deference. Among those who pursue this line, there is no agreement about which normative consideration(s) is, or should be, decisive. It is argued that what matters is whether, inter alia, one has voluntarily taken on the obligations (Simmons, 1979, p. 16); one has been treated fairly (Hart, 1955, p. 185; Dagger, 2000, p. 110); one should be grateful for any benefits received (Walker, 1988); one should act as demanded by a social contract (Darwall, 2011, p. 105); one simply belongs to the community in question (see Horton, 2010, p. 168, p. 175); and so on. These disagreements are not unimportant: they pertain to the questions of who has authority and why, and what the content of the authoritative directives is or should be.

Nonetheless, there is one area of consensus: namely, that what one is obliged to do by an authoritative directive may clash with other reasons for action, and yet one is obliged to act independent of the quality of the action and regardless of one’s view of the balance of reasons in this case (see Hart, 1955, p. 186; 2012 [1961], p. 81; Horton, 2010, p. 13; Knowles 2010, pp. 15–16; Darwall, 2011, p. 105; Simmons, 2016, p. 13, p. 15). As a result, these arguments seem well placed to account for, to explain, the deference at the heart of practical authority, where one is duty bound to obey an authoritative directive. As we shall see, some of those who pursue the norm-based approach nonetheless are moral conflict sceptics, a position that leaves no room for the dilemma of authority. Nonetheless, a norm-based approach is preferable, I will try to show, for the simple reason it is at least compatible with an argument acknowledging the dilemma of authority.

One version of the norm-based approach is voluntarism. It states that one’s obligations are justified if one has rendered oneself obligated by one’s own voluntary acts (Plamenatz, 1968 [1938], p. 154; Simmons, 1979, p. 16). For instance, if one has promised to obey a regime, one is duty bound to do so because its rules have the required normative justification. Perhaps there are some things one should never promise to do, including when the promised action is judged immoral from an independent standpoint. Even so, it can happen that one is bound by a promise, and yet there may be reasons not to act as promised, including because the required action, or its consequences, would be ‘grossly unjust’ (Warnock, 1971, pp. 115–16). The basic idea here is that a promise gives rise to an obligation simply because the promise is voluntary, and independent of the quality of the action promised (Hart, 1955, p. 184; 2012 [1961], p. 197). This is clearly different to the reason-conforming approach. Here, the voluntariness of one’s promise is sufficient to render one obligated: it is not also necessary that being bound in this way better ensures one’s conformity with reason. As we saw, one may be obliged to obey a directive that is, in some way, irrational and immoral, for example, an unjust law.

Some will go further, arguing that a promise to obey a regime is both necessary and sufficient for legitimacy: i.e., one cannot be obliged in any other way (see Pateman, 1989, p. 60). Whether or not that is the case is beyond the scope of this paper. Instead, I simply want to focus on the suggestion that a norm-based approach, whichever approach it is, is well placed to account for the deference demanded by authority. And this seems to be the case, because it is an approach which presupposes that the legitimacy of authority is one thing, and what the weight of reason demands another. This is true, for example, of the social contract version of this argument. The idea here is that one only has obligations to a just regime, as only a just regime would be chosen in an original contract, but also, one may be obliged to obey unjust laws passed by a just regime, so long as they do not ‘exceed certain limits of injustice’ (see Rawls, 1971, p. 343, p. 351).

How could a just regime pass unjust laws? Such a scenario is not difficult to imagine. Let us say a rule has been introduced in higher education stipulating that each University help bring about what John Rawls calls ‘fair equality of opportunity’: namely ‘equal prospects for culture and achievement for everyone similarly motivated and endowed’ (Rawls, 1971, p. 73). This is a just rule. It aims to ensure people are rewarded in accordance with their natural endowments (combined with the efforts they make). And yet there are reasons to object that the new rule is, also, unjust. For it does not address inequality in natural endowment, including natural disabilities. If the new rule has legitimacy, it should be obeyed. This is the case regardless of how weighty one’s reasons are to instead strive to overcome inequality in natural endowment. This example illustrates how what one is obliged to do cannot be conflated with what morally speaking one ought to do. Yes, there are normative reasons justifying authority, but a legitimate authority may oblige one to act in ways that are judged immoral when the merits of the case are considered. In such a scenario, one must decide whether the harm caused by obeying unjust laws outweighs the duty to obey those laws (Simmons, 1979, pp. 78–9; 1999, p. 769; 2005, p. 148).

Of course, I have barely begun to examine the many different norm-based approaches to authority (see Fives 2022a; 2022b). But, the little we have seen suggests this approach is well placed to make sense of the idea that, if one is duty bound to a legitimate regime that has passed a law that is in some way unacceptable, one is obliged to obey, regardless of the quality of the obligatory action or the weight of some of the reasons not to act as obliged. Because it is assumed the justification of deference is one thing, and the weight of reasons (including moral reasons) another, it is at least possible that the two can be in conflict. When the two are in conflict, and when the defeated reason is not cancelled, then, no matter what one does, one should feel regret for the wrong one does.

I also warned that the norm-based approach can be taken in a different direction, one that is sceptical of moral conflict (Sect. 1). Some have hoped to show that, with respect to one significant moral commitment, there should be no conflict between the demands of authority and the weight of reasons. Just one exampleFootnote 4 is Stephen Darwall’s thesis that the following ideas are, as he says, ‘interdefinable’: namely, ‘practical authority’, ‘accountability to’, ‘second-personal reason’, and ‘valid claim and demand’. His thesis is that, if practical authority is derived from a social contract, it follows, one is accountable to that authority and to the claims it makes (i.e., they are second-personal reasons), but also, the claims they make are valid. Indeed, in Darwall’s work, a valid claim or demand is ‘one that is within the authority of someone having practical authority with respect to another to make’ (2011, p. 110). Insofar as authority has the requisite normative justification, deference to its authoritative directives is valid. That might be so. However, if Darwall’s intention is to remove the spectre of unresolved moral conflict, he has not succeeded. The argument here does not show there is no unresolved dilemma of authority. Granted, when authority has the requisite normative justification, one ought to defer; and also, any directive issued by such an authority will be a valid reason for action. But this establishes nothing more than the tautological truth that an authoritative directive is ‘valid’ qua authoritative directive. Nothing follows about whether it is also valid qua reason more generally.

Indeed, in many of the arguments advanced as part of the norm-based approach it is clear that the dilemma of authority can leave regret. For instance, in their respective arguments, both Herbert Hart and Alan John Simmons maintain that freedom is to the fore in justifying authority. Hence, authoritative restrictions on liberty are justified insofar as they are either necessary to ensure respect for the autonomy of all (Simmons, 1979, p. 45; 2005, p. 149; 2016, p. 122), or invoke the equal right of all to be free (Hart, 1955, p. 188). Despite this, they each concede that authority can come at a cost to liberty (Hart, 1955, p. 182; Simmons, 1979, p. 65; 1999, p. 746). Both also note that, when a legitimate rule is in conflict with other reasons for action, it is an open question how such a conflict is to be resolved (Hart, 1955, p. 186; Simmons, 1979, p. 8; 2016, p. 130, p. 131).

It is an open question how such a conflict is resolved, and as I said, I do not try to answer that question here. My argument is that, regardless of how the dilemma of authority is resolved, it might be that the defeated reason is not also cancelled. When the conflict is resolved, it can leave regret. This is evident when we consider the clash between authority and freedom, whenever the two are in conflict. That is, the requirement to protect individual freedom is a reason for action. But as a reason it can conflict with the requirement to defer to a legitimate authority. If the requirement to protect individual freedom is defeated but is still binding, it is rational to feel regret for failing to act as that reason demands.

We have seen that even those who start from a commitment to individual freedom concede that such a conflict can arise. Indeed, in extreme situations, the liberty of the few is sacrificed for the sake of the system that guarantees liberty for all. Rawls explores this issue through the following example. Armed bands are being formed ‘in preparation for civil strife’, he says, and the threat is so grave it becomes necessary to convict some people for concealing firearms on their property even when this has been done without their knowledge (Rawls, 1971, p. 242). Any other approach risks letting the extremists off the hook, leaving them free to threaten the liberty of all. In this example, it is assumed the law has legitimacy: it is based on a social contract it is rational for all to accept. If that is conceded, there is a duty to obey, and yet it requires acting in ways that are also wrong, where it is rational to feel regret. Could one fail to feel regret and still claim to be rational, and moral? Not if the requirement to protect individual freedom is a reason for action; and not if one wishes to avoid sinking into moral indifference. This is a situation where an authoritative directive is justified, but weighty reasons conflict with it. Perhaps the directive is outweighed, but even so, it is not cancelled because it still has legitimacy. One should defer, but also one should not defer. Such conflicts are genuine dilemmas, where, no matter what one does, one does something it is rational to regret. And such dilemmas are a possibility if the justification of authority is a separate matter from the demand that one freely act on the weight of reasons.

6 Conclusions

When the dilemma of authority arises the demands placed upon one are incompatible. This is the case when one ought to defer to authority, one ought to meet the demands of freedom and rationality, one can do either, but one cannot do both. The reason-conforming approach has promised to show why one better satisfies the demands of freedom and rationality by accepting authority when it is legitimate, but it has done so in such a way as to fail to account for deference. In contrast, an approach that instead starts from the normative bases of legitimacy just might be able to account for the deference at the heart of authority. More to the point, it might succeed in accounting for the dilemma of authority because it accepts that the demand for deference is one thing, the weight of reasons something else.

What next, if we are faced with an unresolvable dilemma? I do not think the answer is the rejection of authority. This is because authoritative directives are a part of everyday moral experience. But nor is complacency justified. For it is precisely when authority has legitimacy that a conflict can arise between two types of reason, and two kinds of reasoning. Those who do obey a legitimate authoritative directive are acting for a reason (i.e., a content-independent, exclusionary reason); but are also failing to act for a reason (i.e., the strength of the first-order reasons that are available).

Therefore, authority is not simply a moral problem. It is also a dilemma. And what we have seen is that efforts to escape from the horns of this dilemma fail. Just because an authoritative directive has legitimacy it does not follow that in deferring there is no loss to freedom and rationality. Indeed, it is only possible to ward off any losses to freedom and rationality by inadvertently doing away with deference itself. This suggests that the dilemma of authority involves a choice, indeed a tragic choice, between independent and conflicting moral considerations. The choice is tragic because authority, when legitimate, comes at a cost, the irredeemable cost to freedom and rationality.