Virtue Ethics and the Morality System

Abstract

Virtue ethics is frequently billed as a remedy to the problems of deontological and consequentialist ethics that Bernard Williams identified in his critique of “the morality system.” But how far can virtue ethics be relied upon to avoid these problems? What does Williams’s critique of the morality system mean for virtue ethics? To answer this question, we offer a more principled characterisation of the defining features of the morality system in terms of its organising ambition—to shelter life against luck. This reveals the system to be multiply realisable: the same function can be served by substantively different but functionally equivalent ideas. After identifying four requirements that ethical thought must meet to function as a morality system, we show that they can also be met by certain constellations of virtue-ethical ideas, including notably Stoicism. We thereby demonstrate the possibility of virtue-ethical morality systems raising problems analogous to those besetting their deontological and consequentialist counterparts. This not only widens the scope of Williams’s critique and brings out the cautionary aspect of his legacy for virtue ethics; it also offers contemporary virtue ethicists a more principled understanding of the functional features that mark out morality systems and lie at the root of their problems, thereby helping them avoid or overcome these problems.

1 Introduction

The renaissance of virtue ethics is driven by a dissatisfaction with deontological and consequentialist ethics and by the hope that the ideas of the ancients can offer us attractive alternatives. The work of Bernard Williams is emblematic in this connection, for he spearheaded both the critique of modern ethical theories and the return to ancient ethical thought.¹ He criticised “the morality system” he found exemplified in Kantianism and (to a lesser extent) in utilitarianism; and he argued that ancient ethical thought could offer attractive alternatives. In Shame and Necessity (1993; henceforth: SN), he focused on the Homeric Greeks and the picture of ethical life that emerges from tragedy. But in Ethics and the Limits of Philosophy (2011; henceforth: ELP), he claimed, more broadly, that “very old philosophies may have more to offer than moderately new ones” (ELP 220), and expressed the hope that the concept of virtue, in particular, would “come back into respectable use” (ELP 10). According to ELP, we should look to virtue-ethical ideas for a salutary alternative to the morality system.

Yet Williams remained noncommittal about what the characteristic features of that system were, defining it mainly by example: its “purest, deepest, and most thorough representation” is Kant’s ethical theory (ELP 193–4). In a later essay, however, he identified the characteristic organising ambition of the system: “to provide a shelter against luck” (1995c, 241). The system promises to shelter life from luck by ensuring that opportunities to achieve what really matters are distributed on an ultimately fair basis.² But as Williams highlighted, recruiting and warping ideas to provide a shelter from luck brings four problems in its wake: (i) the problem of demandingness: agents have to sacrifice too much in order to comply with the demands of the system³; (ii) the problem of integrity: the system leaves no-one in particular for individuals to be, alienating them from the personal projects which sustain the possibility of a meaningful life; (iii) the problem of the reductive view of ethical experience: by focusing on the significance of a handful of considerations to the exclusion of all other types of considerations, the system leaves too few ethical resources to be true to lived ethical experience; and (iv) the problem of the irredeemable promise: due to its lack of psychological realism, “the aim of making morality immune to luck is bound to be disappointed” (Williams 1981a, 21), so that the promise must in the end prove illusory.⁴

These problems will be spelled out in §4; but our aim is not to adjudicate whether they prove fatal to deontological and consequentialist ethics—a large question on which there is a substantial literature.⁵ The question we pursue arises once we grant that Williams identified, if not insuperable hurdles, then at least real shortcomings of the morality system. Mining virtue-ethical ideas as yet untouched by Christianity then promises to offer ways of circumventing the problems Williams identified. But to what extent are virtue-ethical ideas themselves free of these problems? Could the shortcomings of the morality system also be unwittingly recreated using virtue-ethical material?

Our guiding idea is that once the system’s organising ambition to provide a shelter from luck is recognised, one sees not only why Kantianism—and the Kantian aspects of utilitarianism, such as its strong underlying sense of moral obligation (ELP 198)—constitutes a realisation of the system, but also that this is only one way in which ethical ideas can be arranged to hold out the promise of sheltering life from luck: once characterised in functional rather than substantive terms, the system is seen to be multiply realisable. The implication is that functionally analogous but substantively different systems could be discerned in the history of philosophy or come out of contemporary philosophy. Here, we argue that virtue ethics can indeed generate its own virtue-ethical morality systems, and these attract Williams’s critique just as much as modern deontological and consequentialist theories do.

Our argument has a historical and a systematic upshot. Historically, we show, using more precise criteria than were hitherto available, that morality systems in Williams’s sense go back to ancient ethics.⁶ Systematically, we show that the morality system has many faces, and that even contemporary virtue ethicists can end up recreating the functional characteristics of the system despite steering clear of deontology and consequentialism. To protect against this risk—or even just to overcome these problems if, pace Williams, they are taken to be soluble—we offer a more principled understanding of what makes a constellation of ethical ideas function as a morality system: one that is less closely tied to Williams’s Kantian target and allows us to see what would count as recreating features of the system in virtue-ethical guise.

We proceed as follows: in §2, we introduce the claim that what makes the morality system a system is the ambition to provide a shelter from luck. This yields a functional characterisation of the system which we illustrate using Kantianism, thereby also showing how it makes sense of why Kantianism combines the features it does. In §3, we then use this functional characterisation to identify functionally analogous constellations in Greek thought; we focus on Stoic ethics as the most paradigmatic example and only hint at its continuities with the Socratic-Platonic tradition; of course, Williams, in highlighting salutary features of ancient ethics, had in mind Greek thought that was earlier than this; but our aim in focusing on the Stoics is to bring out how a shelter from luck can not only be constructed using the modern material that Williams identified, but also using the very different ideas of virtue (arete) and well-being (eudaimonia). In §4, we consider what implications this has for contemporary ethics. Finally, in §5, we contrast our findings with those of philosophers who see substantive parallels between ancient ethics and modern morality. We argue that it is precisely the differences that make it worthwhile to see how virtue-ethical ideas can form a morality system of their own, because this alerts us to sui generis ways in which virtue ethics, though billed as a liberating alternative to deontological and consequentialist ethics, can recreate the same problems in a different guise.

2 The Functional Characterisation of the Morality System

In ELP, Williams is clear that Kant’s ethical theory, and the Kantian aspects of utilitarianism, are in the crosshairs.⁷ From this one can derive a substantive characterisation of the system. But Williams insists that the system “is not one determinate set of ethical thoughts,” but rather “embraces a range of ethical outlooks”; “the important thing about [it] is its spirit, its underlying aims, and the general picture of ethical life it implies” (ELP 193). This invites us to look for a more abstract characterisation of the system that allows us to specify the “underlying aims” that Kantianism serves.

Williams suggests that the system grows out of a longing for “ultimate justice” (ELP 43). It is, at base, a reaction to the fact that “most advantages and admired characteristics are distributed in ways that, if not unjust, are at any rate not just, and some people are simply luckier than others” (ELP 217). The talents and dispositions one is born with, the resources one can draw on, and the opportunities one is presented with, are all ultimately a matter of contingent empirical determination. This fundamental lack of justice or fairness in the distribution of advantages is the root problem the system seeks to remedy. Promising to lift people out of this dispiriting condition, the system expresses “the ideal that human existence can be ultimately just”; it offers a way of living well that “transcends luck” (ELP 217). This is the system’s allure. It promises “solace to a sense of the world’s unfairness” (1981a, 21). As Williams puts it in a later essay,

the point of this conception of morality is, in part, to provide a shelter against luck, one realm of value (indeed, of supreme value) that is defended against contingency. (1995c, 241)

This allows us to characterise the system in functional terms, by reference to a salient beneficial effect that living by that system is supposed to have: to provide a shelter from luck. Williams’s formulation (‘in part’) allows that this may not be all it does—it may, in part, serve other points as well. But Williams never mentions any other point it might serve, and clearly, the salient point of the system, for Williams, is to shelter life from luck.⁸

Once characterised in such functional terms, however, the system is seen to be multiply realisable, so that there could be other versions of it besides the deontological and consequentialist ones. We can abstract away from the peculiarities of its Kantian or utilitarian realisations and identify a functional template describing the various subsidiary tasks that need to be discharged for the system to have the capacity to shelter us from luck.

In order to provide a shelter from luck, the system pursues a four-pronged strategy: (i) it introduces a special kind of good or value; (ii) it presents that special good or value as supreme, not just in the sense of being worth more, on balance, than anything else, but in the sense of drowning out any rival consideration by being the only good that truly matters; (iii) it puts the capacity to realise this supreme good or value entirely within our control; (iv) it ensures that the perspective from which the special good appears supreme should always apply and always prove decisive.

Let us examine not only how Kantianism implements this strategy, but also how we can derive from these implementations a more general sense of the functional demands one must meet in constructing a shelter from luck.

First, the system has to identify a special kind of goodness or value that it treats as being of a radically different order. In Williams’s example, this is moral value, which contrasts with the value of aesthetic merit, athletic excellence, or admirable qualities such as wit, talent, and intellectual brilliance. The deep contrast that the system establishes here goes beyond merely acknowledging some distinction between the moral and the non-moral. Hume, for example, distinguishes between character traits commonly hailed as moral virtues and mere talents, but he thinks the distinction marks no deep contrast, and is merely “grammatical.”⁹ Similarly, Williams thinks that some moral/non-moral distinction is likely to arise as a way to encourage behaviour which takes the interests of others into account.¹⁰ The system’s elaboration of the moral/non-moral distinction, however, has to fashion it into a starker contrast, because the system needs to emphasise how fundamentally different moral value or goodness is from anything else.

Second, this special kind of good or value has to be the only thing that ultimately matters if it is to truly compensate for lack of luck in all other respects. It cannot be “merely a consolation prize” (ELP 217). The relevant good has to be supreme—not just in the sense of being worth more, on balance, than anything else, but in the sense of drowning out any rival consideration by being the only good that truly matters. In Kantianism, this is the goodness of the will to do one’s duty because it is one’s duty. But at an abstract level, the functional requirement is that the good must be an aspect of life such that if life displays that aspect, that is all that matters—at least from the perspective of the system.

Third, the capacity to pursue this supreme good must be completely within the agent’s control—for otherwise, it would once again be a matter of luck whether one found it more or less easy to make life go well. What determines whether life goes well cannot be dependent on how the consequences of one’s actions play out as a result of contingent forces beyond one’s control. To offer an effective shelter from luck, whatever it is that counts towards the realisation of the supreme good in the system must retreat into the agent. It must lie in trying rather than succeeding. But even within the agent, it must lie beyond empirical determination, “in a kind of trying that lies beyond the level at which the capacity to try can itself be a matter of luck” (ELP 217).

Kant’s ethical theory fulfils this requirement to the letter: it insists, first, that moral goodness pertains to the good will rather than to its consequences; and second, it understands the good will in such a way that one’s capacity to exhibit it remains unconditioned by contingent empirical circumstances: only the will that reflects nothing but our capacity to reason counts as good; the will to do one’s moral duty out of some natural inclination does not count. This is only consistent if the point is to shut out luck. If the achievement of the good could in any way be helped along by natural inclinations, that would once again let luck into the system, since which natural inclinations one finds oneself with is a matter of empirical contingency. Consequently, the will to do one’s duty only counts if it is motivated by reason rather than contingent inclination, for the rational faculty is something which, on Kant’s view, is perfectly evenly distributed. Hence, the only goodness that ultimately matters is the goodness of the unconditioned will of the noumenal self. This distinctive combination of psychology and metaphysics is not necessarily shared by other ethical theories. But it is starkly illustrative of the third requirement on the system’s functionality as a shelter from luck: in order for the capacity to attain or realise the supreme good to be distributed on an ultimately fair basis, this capacity must be entirely within the agent’s control and must not be conditional on anything contingent.

Fourth, for the system to be an effective shelter from luck that protects any agent in any situation, the system needs to provide a perspective that “allows no emigration” (ELP 197): it must be impossible for the normative perspective of the system to fail to apply to agents (because they happen to lack certain motivations, for instance), and when it applies, it must be impossible for the demands of the system not to be what the agent has most reason to do, all things considered. In other words, the system’s demands must apply unconditionally, independently of any motivations that the agent merely happens to have, and the demands must be decisive, so that what the system demands is what the agent really must do. The mere fact that something is the supreme good from the perspective of the system is not enough to meet this requirement. After all, from an aesthetic perspective, achieving beauty might be all that matters; but, unless you believe that the aesthetic perspective is the ultimate perspective, as some nineteenth century aesthetes maintained, this is just one normative perspective alongside others, and hardly always the decisive one. Similarly, it is not enough for morality to be immune to luck; it is only if morality always applies and always makes the most important claim on the agent that morality’s immunity to luck amounts to the agent’s immunity to luck through morality (1981a, 21).

To meet this fourth requirement, the Kantian elaboration of the system develops a special notion of moral obligation that renders the system’s demands categorical. Yet this special notion of obligation that Kant elaborated is not essential to the construction of a shelter from luck. What is essential is that the underlying functional requirement should be fulfilled: the perspective identifying the supreme good should always apply and always prove decisive. We can summarize these two aspects by saying that it should be a binding perspective.

In sum, the system’s capacity to function as a shelter against luck depends on the extent to which it fulfils (i) the special kind of good requirement; (ii) the supreme good requirement, (iii) the control requirement; and (iv) the binding perspective requirement. By fulfilling these requirements, a constellation of ethical ideas renders any lack of luck along other dimensions irrelevant, because it turns the only thing that ultimately matters into something that everyone always must and can realise.

Having identified the spirit and underlying function of the system as being to shelter us from luck in a world that is in many ways not fair, and having identified the four functional requirements that this imposes on the system, we can now ask whether this reveals the system to have more faces than its merely substantive characterisation would lead us to expect. In particular, we can ask whether the ancient tradition of virtue ethics—the main alternative to deontological and consequentialist ethics—harbours its own form of the system.

3 Illustrating the Possibility of Virtue-Ethical Morality Systems

“Impressed by the power of fortune to wreck what looked like the best-shaped life,” Williams writes in ELP, some of the Greeks “sought a rational design of life which would reduce the power of fortune and would be to the greatest possible extent luck-free,” something that “has been, in different forms, an aim of later thought as well” (ELP 5). This is one of the rare passages where Williams hints at how the Greek pursuit of “rational self-sufficiency” and “freedom from the damage of contingency” (2006b, 45) prefigures the aim pursued in different ways by the modern morality system. But focused as he was on criticising the Kantian and utilitarian elaborations of that system, which dominated philosophy when ELP was written, Williams did not explore further how a comparable structure could be formed out of virtue-ethical ideas.

Now that the revival of virtue-ethical ideas that Williams hoped for has materialised, however, it is important to recognise the distinctive way in which virtue-ethical ideas can form a system that is functionally analogous to the modern morality system despite their substantive differences. In this section, we show how even virtue-ethical ideas can form a morality system. We use Stoic ethics as our main case study, since it most vividly displays the functional characteristics of the system¹¹; but insofar as Stoic thought builds on more widely shared Socratic-Platonic ideas, it is to that extent representative of ancient ethics more broadly (though not necessarily of Pre-Socratic thought). We thereby illustrate how a shelter from luck can not only be constructed using the modern material that Williams identified, but also using the very different ideas of virtue (arete) and well-being (eudaimonia).

The first functional requirement on a morality system, the special kind of goodness requirement, is met by the Stoics through their account of what makes life go well. They take contribution to a person’s eudaimonia as their criterion for goodness (see Sext. Emp. Math. XI, 21–6 = LS 60 G), and use it to distinguish between good things (agatha), bad things (kakia), and indifferent things (adiaphora) (see Diog. Laert. VII, 104–5 = LS 58 A). On this basis, they distinguish between the goodness of the virtues and everything else: the Stoics consider only the virtues, which they believe always come in unity—in the form of virtue—to be a beneficial or life-enhancing good to the virtuous person, and the opposite, vice, to be the only bad or harmful thing; all other things, which are contingent, are neither good nor bad (see Diog. Laert. VII, 101–3). These indifferents (adiaphora) notably include death, health, political power, possessions, pain, and social relations. Some indifferents are allowed by some Stoics to have value (axia). But the goodness of virtue is of an altogether different order.

This distinction, which starkly contrasts the goodness of virtue with the indifference of all other things, is the functional equivalent of the modern morality system’s moral/non-moral distinction. Looking to the function of the distinction within the system rather than to the substance of the distinction allows us to sidestep the oft-made observation that Greek ethical thought “lacks words or concepts corresponding at all closely to those of the moral and non-moral” (Annas 1992a, b, 120),¹² and “basically lacks the concept of morality altogether, in the sense of a class of reasons or demands which are vitally different from other kinds of reason or demand” (Williams 2006b, 44).

We can grant all this, and still explore how the distinction that Stoic thought does draw, between the special goodness of virtue and the indifference of everything else, serves the purpose of constructing a shelter from luck. Both the Stoics’ conception of the virtues as consisting in internal goods, i.e. in forms of knowledge, and their belief in the unity of virtue serve the purpose of shutting out luck¹³: since attachments to other people and external goods render us vulnerable to luck, attributing goodness only to virtues that can be achieved within, without such attachments to the external, reduces the extent to which one is at the mercy of fortune; and the doctrine of the unity of virtue further reduces one’s vulnerability to luck by foreclosing the possibility of facing irreducibly conflicting demands. There are no tragic forks on the virtuous path.

This way of conceiving of the goodness of virtue also meets the supreme good requirement. For the Stoics, “that which benefits must be completely superior to that which does not benefit; but nothing is superior to the good” (Clem. Al. Prot. 1.18.63. 1–2 = LS 60 I). In principle, the idea of a supreme good allows for the existence of other goods. Kant, for example, believes that pursuing happiness by following one’s inclinations is a good thing as long as this self-interested pursuit does not interfere with the fulfilment of one’s duty. For the Stoics, however, virtue is supreme because it is the only good. Their view is radical in that the classes of good and bad things contain only a single item each, and everything else is placed in the class of indifferents. All items in this broad class—which is exactly co-extensive with the class of things that depend on luck—are taken to be equally irrelevant for the possession of virtue and the quality of one’s life. Whether you happen to be slowly tortured to death or not is as irrelevant to your well-being as whether the number of hairs on your head is even or odd.

Defenders of Stoicism frequently point to refinements introduced by a further distinction according to which some indifferents have value (axia) and are therefore selected or preferred things (proegmena). This theory of proegmena divides the class of indifferents, which are neither good things (agatha) nor bad things (kakia), into subclasses differing in value (axia). First, there are things that are always and completely indifferent and have no value, like having an odd or even number of hairs. Since it is neither in accordance with nor contrary to nature to have an odd or even number of hairs, there is no reason to prefer one or the other (see Stob. 2.84, 18–85, 11 = LS 58 D). Second, there are things that can be more or less in accordance with nature, and insofar as they are in accordance with nature, they have value, but not goodness: survival, health, strength, or well-functioning organs, but also wealth, good reputation, and noble birth; and some things are indifferent but contrary to nature. Like other animals, humans have impulses (hormai) and reasons to act towards these things which are in accordance with nature (see Stob. 2.79, 18–80, 13; 82, 20–1 = LS 58 C). It is therefore natural and rational for us to desire these things which have value and are selected or preferred (proegmena) and to reject other things (apoproegmena) (see Diog. Laert. VII, 85–6 = LS 57 A, 2).

But even allowing for preferred indifferents with value, the special goodness of virtue remains supreme. Just as “the light of a lamp is obscured and overpowered by the light of the sun,” so the brilliance of virtue obscures and overpowers everything else; there is no “value so great as to be preferred to virtue” (Cic. Fin. III, 44–45). For “in the court the King is not in the rank of the preferred, but they are preferred who rank after him” (Stob. 2.84.18–85, 11 = LS 58 E). The Stoics thus treat the goodness of virtue as lexically prior to the value of preferred indifferents, thereby maintaining a strict hierarchy between them—kingly virtue reigns supreme.

How do the Stoics meet the third functional requirement on a morality system, the control requirement? They provide what Williams called an “ethicised” or ethically motivated description of psychology and agency that “gets its significance from ethical categories” (SN 43): instead of making ethical ideas fit an antecedent description of psychology, they fit their description of psychology to ethical ideas.

The distinction between what lies within our control (eph’ hemin) and what lies beyond it—and is in that sense a matter of luck—is the basis of Stoic ethics. Claiming that virtue is the only good locates goodness in the agent, and thereby goes some way towards meeting the control requirement. But merely focusing on the agent cannot yet guarantee that the relation of virtue and well-being will be sheltered from all contingencies. Too many contingent forces act on the agent—not only external forces, but also the inner forces of passions and impressions.

To achieve further protection from contingency, virtue must retreat further into the agent.¹⁴ This is where the highly intellectualistic nature of Stoic psychology, and in particular of the Stoic model of agency, comes in. Its intellectualism lies in its claim that everything that matters is only a matter of judgement, and judgment is the only thing that is fully under control of the rational and leading faculty of soul, the hegemonikon. Marcus Aurelius aptly calls this faculty of soul the “inner citadel,”¹⁵ because it is shielded against external forces.

A number of Stoic theorems serve to fortify the inner citadel. The contingent influences of the passions, for example, which Plato sought to isolate from the “rational concerns that aim at the good” (SN 43) using his tripartite model of the soul, are not thought of by the Stoics as something outside the inner citadel of reason at all. Rather, they are internalised in a way that tames their contingency: the passions are described as modifications of reason based on judgements such as “This is bad!” Reason can always assent to or refrain from assenting to passions, thus keeping the passions fully within its control.¹⁶

Reason’s gatekeeping powers also extend to impressions—be they impressions from the senses (e.g. hearing something) or from the activities of the mind (e.g. dreaming or calculating) (Diog. Laert. VII, 49–51). Contingencies may affect which impressions we experience, but they cannot affect reason’s judgements about them. The hegemonikon has the critical power of giving or withholding assent (synkatathesis) to every impression. Reason’s judgment also controls which impressions result in action. On the Stoic model of agency, an action (prattein) is understood in terms of the impulse (horme) out of which it arises, i.e. in terms of the movement of the soul towards something.¹⁷ First and basic impulses are natural, growing out of an innate awareness of physical constitution and functions (see Sen. Ep. 121, 6–15 = LS 57 B)—they make animals and humans act in a way that is appropriate (oikeion) to their natural and particular constitution. But what of impulses triggered by impressions? If an impression arising out of contingent circumstances, such as the conjunction of hunger with the presence of food, led immediately to an impulse to eat and to actual eating, humans would lack full control over their actions. The Stoics remedy this by insisting that unlike other animals, humans do have full control because there is a (literally) decisive intermediate step: only if the hegemonikon decides to assent to an impression does that impression become an impulse issuing in action. To take a modern example: sitting in a dentist’s chair and receiving an impression of pain as a result of the treatment, one can sometimes decide not to assent to this impression and remain unflinchingly seated. According to the Stoics, it is always like that: reason decides which impressions turn into impulses.

Once he has assented to an impression and his impulse to act is in effect, however, even the Stoic Sage, who embodies the perfectly virtuous person, cannot expect all his actions to be successful, because even he cannot control how things unfold (see Sen. Ben. 4, 33 or Arr. Epict. diss. 2, 6, 9; 2, 16, 15). The Stoics have two strategies to deal with this problem, one ex ante and one ex post.

The ex ante strategy is to maintain that the goodness of virtue cannot be spoiled by contingent events preventing the action’s successful execution, because virtue is complete before the virtuous action can be interfered with by contingency. Virtue depends on the soul’s inner activities rather than on the success of external action. Controlled by reason, these inner activities remain unaffected by external contingencies (Diog. Laert. VII, 128). Judgement and impulse are qualified by the famous “reverse clause” (hupexhairesis, exceptio), which can also be rendered as: “if nothing prevents,” “if fate permits,” or “God willing.”

The ex post strategy concerns the Sage’s behaviour and decision once his virtuous action has been interfered with by contingency. The Sage’s “mind adapts and converts everything that impedes its activities into something that advances its purposes, and a hindrance in action becomes an aid” (M. Aur. Med. 5, 20, see also Arr. Epict. diss. 4, 1, 100–102).

It is thus clear that on the Stoic picture of psychology, the control requirement can be met: the distinction between what is within and beyond our control aligns with that between what is supremely good and what is merely indifferent.

When it comes to the binding perspective requirement, finally, the Stoics seem ill-equipped to meet it, since they lack the “special notion of moral obligation” (ELP 202) that is key to the Kantian way of meeting this requirement. It would be a distorting anachronism to assume that the Stoic conception of the appropriate behaviour (kathekon) implies anything like the later concept of duty, though Cicero’s translation of kathekon as officium encourages that misunderstanding. The primary sense of kathekon is “acting in accordance with nature.” The only imperative the Stoics possess in lieu of Kant’s categorical imperative is the injunction to “live in agreement with nature” (see Stob. 2.77, 16–27; Diog. Laert. VII, 87–9; Sen. Ep. 76, 9–10).

Through a combination of doctrines, however, the Stoics nevertheless succeed in rendering the perspective of virtue binding. They achieve this partly by dint of the doctrines that meet the first two requirements: if the only good is virtue, then considerations of virtue will encounter no resistance from competing considerations. In later moral theories such as Kant’s, something analogous is achieved through moral rationalism, the view, available in weaker and stronger forms, that moral considerations override or silence all other considerations. But the Stoics have no need to distinguish “moral” from other considerations in order to secure the primacy of virtue, since it follows already from the idea that virtue is the only good. It trivially follows that what is good from the perspective of virtue is what we should do all things considered.

Yet the bindingness of the Stoic ethical perspective is reinforced by being understood to be the point of view of the cosmos itself. In this respect, the Stoics meet the fourth requirement by relying on an ethicised description not just of psychology, but also of nature. To the Stoics, the cosmos is fundamentally rational, because all natural things are determined by the rational principle. To live in agreement with nature is what a natural creature ought to do, because what is natural is appropriate and thus normative. The Stoic doctrine that the good life is the life lived in accordance with nature thus relies on an ethicised conception of natural development. Claims of virtue are thus also inescapable because any human being has most reason to live in accordance with nature, no matter what their contingently acquired concerns and dispositions may be.

It is true that while the Stoics’ ethicised conception of nature raises the prospect of human fulfilment in virtue, thereby increasing virtue’s attraction, their lack of a Kantian notion of moral obligation means that claims of virtue do not possess the same silencing bindingness as claims of moral duty. Yet this difference reflects the Stoics’ reliance on an ethicised conception of nature. If the Stoics can meet the binding perspective requirement with less than Kant’s specially demanding notion of moral obligation, it is because they start with more. Their conception of the cosmos as rational means that the Stoics have no need for the elaborate manoeuvres by which Kant seeks to distil, out of nothing but the pure idea of a rational agent, the unconditional applicability and practical necessity of moral claims. As Williams remarked of Aristotle, the Stoics believe that “the world is written, fundamentally and ultimately, in a script that will tell us a lot about how to be,” whereas Kant tries “to come up with a theory of morality that would deal with the fact of autonomy—that is, the fact that we aren’t told what to do by the way the world is” (1999, 152). On the Kantian conception, the starry heavens above and the moral law within are fundamentally different things, so that nature and morality can pull in different directions. That is why it becomes vital to insist that the starting point for practical deliberation should be to ask what the moral thing to do is. Kant needs to understand claims of morality in a way that will equip them to assert themselves against claims of nature.

But the Stoic conception of virtue does not require human beings to overcome claims of nature at all, because they naturalised reason and ethicised nature. To the Stoics, reason is itself part of the physical world and part of the divine reason that is the rational principle that inhabits and governs all things. Life in accordance with nature then converges with life in accordance with ethical ideas: claims of virtue coincide with claims of nature.

The point of this still very broad-brushed sketch of Stoic ethics is not primarily to make an exegetical claim about the Stoics, however, but to illustrate in principle how even a broadly Stoic elaboration of virtue ethics could meet all four functional requirements on a morality system. It is possible to construct a morality system out of virtue-ethical material. This should lead us to adjust our understanding of the scope of Williams’s critique: it targets not just deontological and consequentialist ethics, but certain forms of virtue ethics as well. And this realisation in turn has implications for contemporary philosophy.

4 Recreating Familiar Problems

The realisation that virtue-ethical ideas can form a morality system of their own is a systematic realisation with implications for contemporary virtue ethics. In reviving virtue-ethical ideas, there is a risk of reconstructing virtue-theoretic versions of the system that raise the same problems as its deontological or consequentialist counterparts. If the system is multiply realisable, it can have more faces than Williams’s critique suggests, and gaining a principled understanding of what functional features exactly give rise to its characteristic problems should be of interest even to those who, pace Williams, think that those problems can be overcome.

The injunction that we should look to the Greeks for ethical ideas that are in better shape than ours must be qualified accordingly: virtue ethics does not necessarily avoid the problems of the morality system, because while it may be unspoiled by some of the historical forces that proved formative for modern ethical thought, it is not, for all that, immune to the problems that Williams urges us to avoid. This helps explain why Williams, though he sees the concept of virtue as an important ethical idea, does not present himself as a virtue theorist, and why he focused on the Homeric Greeks rather than on later Greek schools.¹⁸

So what form exactly do the four characteristic problems of a morality system take in virtue-ethical thought? Drawing on our reconstruction of Stoic thought, we can illustrate how these problems arise in a virtue-ethical morality system.

First, a virtue-ethical morality system can invite its own form of the demandingness problem, i.e. that agents have to sacrifice too much of their well-being to comply with the theory.¹⁹ This is an objection that Susan Wolf (1982) has forcefully amplified: obeying the requirements of a morality system leads to moral sainthood, which entrains grave losses in other respects: the system’s special emphasis on moral goods is apt to come at the expense of non-moral goods like health, pleasure, personal well-being or social relations, which Wolf takes to be necessary parts of a “well-rounded” and good life.

Although this argument was aimed against modern theories, it has even more bite when directed against the Stoic Sage, the godfather of moral saints. The Sage is as morally good as possible by having all the virtues. The Sage “does everything well—that is to say, everything that he does” (Stob. 2.66,14–67,4 = LS 61 G). But in listing the virtues that a moral saint must combine, there comes a point “where one might naturally begin to wonder whether the moral saint isn’t, after all, too good—if not too good for his own good, at least too good for his own well-being” (Wolf 1982, 421). The “nonmoral virtues, as well as many of the interests and personal characteristics that we generally think contribute to a healthy, well-rounded, richly developed character” (Wolf 1982, 421) are bound to be crowded out if all the virtues identified by the Stoic system are fully realised and concentrated in a single individual. In one respect, Stoic ethics is even more demanding and more uncompromising than deontological or consequentialist ethics. Whereas the latter still acknowledge the importance of non-moral goods for personal well-being, the Stoic system denies that any of these things are good at all. For the Stoics, losses that are not losses in virtue cannot really be losses at all. Kantians or utilitarians can at least make sense of demandingness and self-sacrifices, and they can give the agent’s loss in well-being some rational or normative weight. But Stoicism does not allow for that. It robs itself of the conceptual resources even to recognise its own demandingness. And a theory that cannot so much as conceptualise demandingness is ill-equipped to hold it in check.

Second, by urging every agent to live the life of the Sage, the Stoic morality system leaves no-one in particular for the individual to be and thereby generates a form of the integrity problem²⁰: it alienates the individual from the personal projects, convictions, and interests with which the individual is most closely identified, and which sustain the possibility of a meaningful life for that individual.²¹ As Williams puts it, there “can come a point at which it is quite unreasonable for a man to give up, in the name of the impartial good ordering of the world of moral agents, something which is a condition of his having any interest in being around in the world at all” (1981b, 14). Once one considers what is involved in having a character, he argues, one can see that the “omission of character is a condition of [the] ultimate insistence on the demands of impartial morality” (1981b, 14). This is because this insistence amounts to what Williams calls an attack on the agent’s integrity “in the most literal sense” (1973a, 116–117), meaning not just that what the morality system demands fails to fit in with one’s character, but that it threatens one’s having a character in the first place. One’s integrity as an agent, i.e. the fact that various actions are one’s own, depends on the fact that those actions can be seen as flowing from the projects and convictions with which one is most closely identified. Yet if all one’s actions reflect nothing but the impartial demands visible from a perspectiveless “point of view of the universe,” there is no interesting sense in which these still are the actions of a particular person at all.²²

The same obliviousness to character and its role in sustaining a meaningful life is displayed in Stoic ethics. It enjoins each person to give up her point of view and deliberate instead from the point of view of the rational cosmos. Marcus Aurelius, for example, urges himself to abandon the perspective of a separate person: “Everything is good for me that is good for you, o universe, nothing is too early or too late which suits your time” (M. Aur. Med. 4, 23). By asking people to step back from the projects and interests with which they are identified and do whatever the Sage would do, a virtue-ethical morality system alienates people from that which allows them to understand their actions and their life as distinctly their own.

The third problem is that a virtue-ethical mortality system risks encouraging a reductive view of ethical experience. By focusing on the ethical importance of a handful of virtues to the exclusion of all other types of considerations, it arguably leaves itself too few ethical resources to be true to our lived ethical experience.²³ Pressing what it can into the narrowly systematised mould of its virtue-ethical ideas and discarding everything else, it neglects the importance of consequences, duties, principles, emotions, intuitions, and attachments. Meeting the four requirements for forming a shelter from luck then threatens to come at the cost of flattening ethical experience by leaving one unreceptive to many of the ideas and sentiments that give it its fine-structure and imbue it with depth.²⁴

The fourth problem, finally, is that the promise of immunising life against luck is bound to remain unredeemed, because it rests on a psychologically unrealistic picture of agency. If we accept that our ethical theories should be answerable to an adequate psychology, the fact that the psychology presupposed by a theory such as the Stoics’ is unrealistic should put pressure on that theory. This exactly parallels Williams’s objection to the Kantian system, which essentially relies on the possibility of purely voluntary acts presupposing the existence of an utterly unconditioned will in order to hold out the promise of ultimate fairness. But this possibility is not intelligible without Kantian metaphysics, since the “dispositions of morality, however far back they are placed in the direction of motive and intention, are as ‘conditioned’ as anything else” (Williams 1981a, 21). And just as the Kantian morality system faces the challenge that it relies on an unrealistic picture of human agency, the virtue-ethical morality system of the Stoics must confront the fact that its intellectualist psychology is susceptible to being debunked by a more realistic psychology.²⁵

The problem is not just that of being factually wrong about human psychology, moreover. It is that a lack of realism fosters unrealistic expectations that in turn distort our ethical judgement. The Stoics admitted that the Sage was, as a matter of fact, “as rare as a phoenix.”²⁶ Yet the unrealistic expectation that the virtuous person, the Sage, should live unerringly, like a God, led the Stoics to elide all difference between those who fall short of that ideal: there are only sages and fools, and one fool is as bad as another (Stobaeus 2.7.11.g). Likewise, if virtue and vice must be complete at their inception, before external forces come into play, there can be no ethical difference between a guilty impulse and a guilty act, between mens rea and actus reus.²⁷ This virtue-ethical morality system thus violates Owen Flanagan’s Principle of Minimal Psychological Realism, that “when constructing a moral theory or projecting a moral ideal,” one should make sure “that the character, decision processing, and behavior prescribed are possible, or perceived to be possible, for creatures like us” (1991, 32); or, as Samuel Scheffler has more succinctly put it, only a psychologically realistic ethical theory can be a human morality (1992, 7–9).

5 Different Morality Systems

While we have found a morality system in ancient virtue ethics, we have not found an earlier instantiation of the same kind of system as that which Williams found most fully expressed in Kantianism. What we have found, rather, is something that functions like the modern morality system and faces similar objections even though it remains substantively very different. There can be distinctively virtue-ethical morality systems which are not just anticipations of the modern one.

This differentiates our finding from that of someone like Julia Annas, who finds in ancient ethics a morality system along much the same lines as the modern one. The principal hurdles for Annas are that the Greeks do not seem to share two salient characteristics of the modern morality system: the notion of moral reasons as reasons that override or silence non-moral reasons, and the notion of moral obligation. She seeks to overcome these hurdles by arguing that the Greeks did in fact have some notion of overriding moral reasons, and that the notion of moral obligation is not really as central to the modern morality system as Williams made it out to be (Annas 1992a, b).²⁸ Roger Crisp makes an analogous argument. He suggests that “we have more in common, ethically, with Aristotle and the Greeks” (2004, 77) because there is “a general conception of practical necessity or ‘bindingness’ running from the Greeks, through Old English, into the modern day” (2004, 84). Crisp also argues that “modern morality is not as dominated by Kantian ideas as Williams implies” and that “those allegedly ‘Kantian’ ideas, including that of obligation, are there in Homer” (2013, 5n9).

But in seeking, however circumspectly, to assimilate features of ancient ethics to modern ethics, one inevitably runs the risk of downplaying real differences between them. Even if the Greeks had something like our modern notions of moral reasons and moral obligation, these ideas were, to all appearances, not central to ancient ethical thought, as they seem to be easy to miss. If one believes that proponents of ethical theories across the ages are “climbing the same mountain on different sides” (Parfit 2011, 419), one may be more ready to discern intimations of modern ideas in ancient thought. But Williams, for his part, approvingly quotes Collingwood’s comparison of the Oxford realists who would “insist on translating some ancient Greek expression as ‘moral obligation’” to “a man who insisted on translating the Greek word for a trireme as ‘steamship’” (Williams 2006d, 181).²⁹

Two notions that were ascertainably central to ancient ethical thought, however, were the notions of virtue and eudaimonia; and our functional characterisation of the morality system has allowed us to identify a morality system in ancient virtue ethics while acknowledging the ways in which it remains profoundly different from modern ones: while modern morality systems paradigmatically centre around notions of moral reasons and moral obligation, but have little use for notions of virtue and eudaimonia, the virtue-ethical morality system has little use for notions of moral reasons and moral obligation, but centres around the notions of virtue and eudaimonia.

It is easy to think that if the history of philosophy is going to be relevant to contemporary philosophical debates, this must be because our predecessors were climbing the same mountain, so that the study of the history of philosophy discerns in the past some of the same concerns and ideas that figure also in the present. But our discussion illustrates that the opposite can be true. Our argument does precisely not depend on flattening the differences between ancient and modern ethical thought; rather, it is just these differences that make it worthwhile to see how virtue-ethical ideas, far from prefiguring the modern morality system, can form a distinctive morality system of their own. Looking back to ancient ethics can inform contemporary thought by revealing genuine alternatives to it—in our case study of Stoic ethics, alternative ways of fashioning a morality system—and the differences within ancient ethics might also suggest alternative ways of drawing on virtue-ethical ideas without ending up with a morality system.

The implication for contemporary virtue ethicists is that those who hope to avoid some of the problems besetting deontological and consequentialist theories by turning to virtue ethics need to be on the lookout for ways in which virtue-ethical ideas may come to recreate a morality system in a new guise. By organising the characteristics of the morality system in light of the ambition to provide a shelter from luck and characterising it in functional terms, we have tried to arm virtue-theorists with clearer criteria by which to identify the features of ethical thought that are the hallmarks of morality systems. If there is one way of constructing a virtue-ethical morality system, there may be others. This is not to say that modern virtue ethics is bound to repeat the mistakes of ancient virtue ethics. But as we have shown, a morality system can be constructed without notions of moral reasons and moral obligation, so that steering clear of these notions is no guarantee that one will not be pulled, by the concern to shelter life from luck, towards some virtue-ethical version of the morality system. Even those who think that the problems Williams identified can ultimately be overcome will be well served by a principled understanding of what engenders those problems.

The legacy of Williams’s critique of the morality system for virtue ethics is thus two-sided: while it invigorates virtue-ethical thought as a salutary alternative to deontological and consequentialist theories, it also carries cautionary implications, which only emerge once one appreciates the full scope of Williams’s critique. Virtue ethics must be wary of recreating the problems of the morality system in another guise. In view of this possibility, we can actually sympathise with those who, like Samuel Scheffler, doubt that “the proper remedy for the defects of contemporary moral theories is a substantial repudiation of modern moral thought in favour of some conception inspired by the ideas of the ancients” (1992, 11). But while his concern is that we risk losing the best aspects of modern ethical theory, ours is that we risk retaining the worst.