The Skill Model: A Dilemma for Virtue Ethics

According to agent-centered virtue ethics, acting well is not a matter of conforming to agent-independent moral standards, like acting so as to respect humanity or maximize utility. Instead, virtuous agents determine what is called for in their circumstances through good practical reason. This is an attractive view, but it requires a plausible account of how good practical reason works. To that end, some theorists invoke the skill model of virtue, according to which virtue involves essentially the same kind of practical reason as ordinary skills. I contend, however, that ordinary skills provide a plausible and informative model of good practical reason only insofar as they are assessed by agent-independent standards. And so virtue, likewise, must be assessed primarily by agent-independent moral standards, if the skill model is to serve its purpose. I consider how agent-centered virtue ethics might avert this dilemma. But I ultimately suggest proceeding a different way.

Such standards, then, do not determine what is called for; instead, this is primarily a function of virtuous agents, on the agent-centered approach. As John McDowell puts it, "although the point of engaging in ethical reflection still lies in the interest of the question 'How should one live?' that question is necessarily approached via the notion of a virtuous person. A conception of right conduct is grasped, as it were, from the inside out" (McDowell 1979: 331). Call this defining feature of agent-centered virtue ethics the primacy of agent assessment.
This feature not only makes agent-centered virtue ethics distinct from act-centered approaches to ethical theory, like deontology and consequentialism (see Watson 1990); it also accounts for much of the distinctive appeal of the approach. Acting well, according to agent-centered virtue ethics, is not a matter of rigorous adherence to supposedly universal moral standards. 2 Rather, virtue, developed through continuous practice and reflection, enables agents to make wise judgments about what to do in their particular circumstances.
But the primacy of agent assessment, so construed, raises a pressing question: if virtuous agency is not a matter of adhering to independent moral standards, then how do virtuous agents reliably determine what to do? Agent-centered virtue ethics require an explicit account of how good practical reason is supposed to work. Moreover, this account must avoid familiar objections, along empirical as well as conceptual lines. Some have argued, for instance, that it is not realistic for humans to reliably determine what is called for, over time and across situations (e.g. Doris 1998); while others have contended that virtue, so construed, is intellectually elitist and therefore an unsuitable ideal for human beings (e.g. Driver 2001).
The skill model promises just such an account, suggesting that the good practical reason of virtue is essentially no different from that of skills like cooking, riding a bicycle, and speaking French. And if this is so, then the challenges can be readily answered. Such skills are so familiar and ordinary that their attainability by, and suitability for, human beings are beyond suspicion. Thus, suspicion about virtuous reasoning, so conceived, is similarly misplaced.
I contend, however, that the skill model of virtue is in tension with the primacy of agent assessment. In Section 2, I develop the first horn of a dilemma, clarifying what the primacy of agent assessment amounts to and arguing that it is essential to the distinctive appeal of agent-centered virtue ethics. In Section 3, I develop the second horn, arguing that ordinary skills can provide a plausible and informative model for virtuous reasoning only if act assessment is taken to be primary in skilled domains and the moral domain alike. In Section 4, I consider how agent-centered virtue ethics might avert this dilemma. But I conclude by suggesting a different way forward.

2
To delineate the class of theories I am calling agent-centered virtue ethics, it is helpful to begin by contrasting it with alternative approaches to ethical theory which are committed to the primacy of act assessment, meaning that they assess agents in terms of independent standards of action. On a deontological view, for example, virtues might be conceived as "sentiments and habitual attitudes leading us to act on certain principles of right" (Rawls 1971: 383). And on a consequentialist theory, the virtues might include "any mental quality that (systematically) produces utility" (Driver 2016: 110). Both views are committed to the primacy of act assessment, because both assess agents as virtuous (or not) according to agent-independent standards of action, though they posit different standards. 3 Agent-centered ethics, too, come in different varieties. On Rosalind Hursthouse's influential view, for instance, an act is right "iff it is what a virtuous agent would characteristi-cally…do in the circumstances" (Hursthouse 1999: 28). Compare this to Michael Slote's claim that right acts are those which "exhibit, express, or reflect (admirable inner) states" (Slote 2001: 17). Both assert the primacy of agent assessment, since both assess acts as right (or not) according to the goodness (or badness) of agents. This feature makes them both agent-centered theories, but only Hursthouse's is an example of what I am calling agent-centered virtue ethics. So, identifying the key difference between them will further clarify what is distinctive about this family of theories.
Slote distinguishes between agent-based ethical theories, like his own, and agent-prior ones, like Hursthouse's. An agent-prior view "treats the evaluation of actions as derivative from independent aretaic character evaluations;" yet "character evaluations are not regarded as fundamental" on such a view (Slote 2001: 6). By contrast, an agent-based view "treats the moral or ethical status of acts as entirely derivative from independent and fundamental aretaic…characterizations of motives, character traits, or individuals" (5). 4 Hursthouse evidently concurs that her view, along with other paradigmatic virtue-ethical theories, is agent-prior rather than agent-based: "A common misunderstanding of virtue ethics (is that) in being agent-centred rather than act-centred, in starting with the virtues and vices rather than right or wrong acts, is committed to a sort of reductionism" (Hursthouse 1995: 72). Along the same lines, Julia Annas explains: The virtues are what may be called primary, as opposed to basic. These are the notions that we start from; they set up the framework of the theory, and we introduce and understand the other notions in terms of them. They are thus primary for understanding...However they are not basic in the modern sense: other concepts are not derived from them, still less reduced to them. (Annas 1993: 9) So, right action is not simply derivative from, or entirely reducible to, virtue. But act assessment still depends upon agent assessment in an important way, on this theoretic approach. As Daniel Russell succinctly puts it: "right action cannot be fully understood without an account of the virtues, whereas the virtues can be understood without an account of right action" (Russell 2009: 69). This formulation comes close to capturing what I am calling the primacy of agent assessment, which I take to be the feature that distinguishes agentcentered virtue ethics from other kinds of agent-centered theories. Just one further clarification is needed.
3 Calling deontological and consequentialist ethics "act-centered" and attributing the "primacy of act assessment" to them is potentially misleading since neither necessarily conceives of virtue specifically in terms of right action. They might instead understand both virtue and right action in terms of a more basic concept, such as right reasons or good outcomes (I thank an anonymous referee for pointing this out). Such theories, however, may still be thought of as act-centered in that they take the standards for action to be independent of, and (typically) prior to, agent assessment. 4 In being agent-based, Slote's view is also subject to my critique. I focus on merely agent-centered theories in order to show that their weaker commitment to the primacy of agent-assessment does not allow them to escape this critique.
A theory might take agent assessment to be primary over act assessment in two different senses, since assessment of acts might depend upon assessment of agents in two different ways. Sean McAleer helpfully distinguishes between the "constitutive" and "criterial" versions of the primacy of agent assessment (McAleer 2007: 220). On the constitutive version of the claim, an act is right because a good agent does or would do it. That is, the goodness of (actual or counterfactual) agents makes right acts right. On the criterial version, the best or only evidence that an act is right is that a good agent does or would do it.
Crucially, if agent-centered virtue ethics posit the primacy of agent assessment only as a criterial claim, then they lose much of their distinctive appeal. This is because act-centered ethics, like deontology and consequentialism, can also posit the criterial primacy of agent assessment. Indeed, given the complexity of moral life, we must often look to moral exemplars, or rely on our own good character, to determine what to do. But this is compatible with supposing that independent moral standards make right acts right. Action guidance can come apart from action assessment in ethical theory.
Notably, proponents of agent-centered virtue ethics often criticize act-centered ethics for prescribing decision procedures which "mechanically tell us what to do" (Annas 2011: 34) by way of a "set of rules or principles that can provide specific action guidance" (Hursthouse 1999: 25) which "anyone can apply…correctly, regardless of their character" (Russell 2009: 60). But there is nothing about act-centered ethics per se that commits them to this implausible view of good moral agency.
For example, act utilitarianism will assess an act as right because it maximizes utility; but it does not follow from this that right action is guided by hedonic calculation. Rather, act utilitarians can recognize that reliably performing right (i.e. utilitymaximizing) acts requires inculcating certain dispositions to respond well to one's circumstances in more immediate ways. 5 Thus, they can posit the criterial primacy of agent assessment: the best or only evidence that a particular act is right is that it is, or would be, done by a virtuous agent. But they will still reject the constitutive primacy of agent assessment, insisting that what makes an act right is the agent-independent fact that it maximizes utility.
In order for agent-centered virtue ethics to provide an alternative to rival views on this point, then, they must posit the constitutive primacy of agent assessment: right acts are right because virtuous agents (would) do them. 6 Importantly, though, this still does not imply that act assessment is entirely derivative from agent assessment or that agent assessment is normatively fundamental, as would be the case for an agent-based theory like Slote's (2001). Again, agent assessment is primary for agent-centered virtue ethics in that it explains the moral status of acts. But it need not be the only, or most fundamental, factor in such explanations. 5 See, e.g., Chapter II of Utilitarianism, where Mill insists that maximizing utility can only be achieved by "the general cultivation of nobleness of character" (Mill 2001)." Contemporary theorists have used the same general strategy to develop various consequentialist theories of virtue. See Railton (19841988), Hooker (2000, Driver (2001), andBradley (2005). 6 As Kawall puts it, in his defense of agent-centered virtue ethics as a distinct approach to ethical theory, "actions, states of affairs, and so forth take on a moral status insofar as virtuous individuals in particular will have certain attitudes toward them" (Kawall 2009: 16). Act-centered ethics, by contrast, ascribe independent moral status to actions, states of affairs, and other such considerations.
To illustrate, the wrongness of wanton torture might be explained as follows: the act is wrong because it causes needless suffering; causing needless suffering is wrong because virtuous people are disposed to avoid doing it; and virtuous people are so disposed because this disposition is necessary for one to flourish as a human being. A virtuous disposition explains the wrongness of the act here, in the italicized clause. But the consequences of the act also factor into the explanation, and the virtuous disposition is itself explained by the more fundamental notion of human flourishing. In this way, agent-centered virtue ethics can posit the constitutive primacy of agent assessment without reducing act assessment entirely or fundamentally to agent assessment, thus avoiding the radical agent-based approach. 7 However, agent-centered virtue ethics must at least commit to the claim that virtue explains the wrongness of acts like wanton torture and not vice versa. Otherwise, they would be merely agent-focused, meaning that their "focus is on the virtuous individual and on those traits, dispositions, and motives that qualify her as being virtuous" (Slote 2001: 4). This weaker approach allows that the moral status of an act might be accounted for entirely by agent-independent considerations. And this would undercut the distinctive appeal of agent-centered virtue ethics, namely, its insistence that rightness is not just a matter of adherence to moral rules, principles, or other such standards.
Now, it might be objected that this analysis distorts agent-centered virtue ethics, since such theories "will be interested in virtuous action, but will not get much out of the notion of right action" (Annas 2011: 47). Accordingly, they will give an account of right action "under pressure, only in order to maintain a fruitful dialogue with the overwhelming majority of modern moral philosophers for whom 'right action' is the natural phrase" (Hursthouse 1999: 69). Moreover, if virtue is understood in terms of the kind of character needed to flourish as a human being, and if it functions to specify what to do in particular cases, then acting rightly is just a matter of acting virtuously, and acting virtuously instantiates the flourishing of the agent. 8 Virtue, so understood, is a criterion of right action. Yet it is not a mere means to satisfying independent moral standards, like respecting humanity or maximizing utility. It functions both to specify what is called for and to instantiate the value of doing what is called for in any given situation. 9 This, however, is not really an objection to the forgoing analysis of agent-centered virtue ethics, but rather an elaboration of it. While virtue is indeed a criterion of right action on such a view, in that it picks out the thing to do in the circumstances, the fact that it does so by bringing the action in line with the agent's flourishing means that virtue also makes it the thing to do, by incorporating it into the good life of the virtuous agent. Thus, virtue is not just a criterion of right action, on such a view. It is constitutive of right action as well. Conceptualizing right action as virtuous action, and understanding virtue in terms of the 7 See Russell (2008) and Kawall (2009) for similar arguments. 8 Thus, one's own flourishing can be understood to impose an "unconditioned condition on any more particular recommendations" about how one ought to act (Baril 2014: 25). See also McDowell (1980), Hursthouse (1999), Foot (2001), Russell (2009), Annas (2011), Bloomfield (2014), and Stichter (2018) for different articulations of this sort of view. 9 Russell (2009) develops an especially thorough virtue-centered account of right action along these lines. On his view, practical wisdom unites all right acts by bringing them in line with the virtuous agent's flourishing. I argue in the following sections, however, that the skill model of virtue does not support this conception of practical wisdom. agent's flourishing, therefore specifies how act assessment can depend upon agent assessment. And while this makes for a plausible account of the primacy of agent assessment, it does not alter the nature of this defining feature of agent-centered virtue ethics. 10 By contrast, if virtue were understood primarily in agent-independent terms-say, as the disposition to act in a way that maximizes utility-then virtue would be, at most, merely a criterion of right action. The fact that an act is, or would be, done by a virtuous agent might still be the best or only evidence of its rightness, especially in hard cases. But virtue could nevertheless be explained completely in terms of independent standards of action, the mark of an act-centered ethical theory. Hence, for agent-centered virtue ethics to offer a distinct and attractive approach to ethical theory, one which makes right action primarily a matter of what virtuous agents (would) do in the circumstances, they must posit the primacy of agent assessment in the constitutive sense, as making right acts right. This comes at the cost, however, of forgoing critical resources offered by the skill model of virtue.

3
Recall that the skill model of virtue promises a plausible and informative account of how virtuous agents determine what to do, which, according to agent-centered virtue ethics, is not primarily a matter of adhering to agent-independent moral standards. Paul Bloomfield explains why the skill model seems well-suited for this task: Beginners follow rules differently than experts, and we see this as being the case in both virtues and skills. As beginners, we learn first by learning general rules and principles, but upon gaining a deeper appreciation for the field we are learning, and for the principles of the field themselves, we learn to see that in particular cases a general rule must be broken. Part of what makes an expert an expert is knowing when it is correct to break the rules... (Bloomfield 2000: 26) In the same vein, Annas claims that, while employing rules principles may be necessarand y for learning certain techniques, "Being versed in technical matters does not bring with it any move to understanding, self-direction, and improvement; that comes only with the development of skill" (Annas 2011: 33). Like practical skills, then, virtue is supposed to involve an understanding of how to act well, developed through practice, which goes beyond mere adherence to rules and other agent-independent standards of conduct.
This conception of good practical reason does apparently apply to many familiar, ordinary skills. A good cook does not simply follow recipes; a good mechanic does not just rely on manuals; and a good musician does much more than play the notes off the sheet. Conceiving of virtue as akin to such skills, then, would indeed seem to offer a compelling picture of good practical reason, and one which is amenable to agent-centered virtue ethics. In the case of both skills and virtue, it appears that the thing to do is primarily a matter 10 I do not argue, then, that agent-centered virtue ethics are committed to a vacuously circular conception of right action or that this approach to ethical theory is otherwise untenable. Notably, the question of whether virtue ethics can give a distinct and plausible account of right action is a longstanding issue: see, e.g., Watson (1990), Stohr & Wellman (2002), Das (2003), Copp & Sobel (2004), Svensson (2011), andCrisp (2015). For my part, I grant that agent-centered virtue ethics can offer a rich and interesting account of right action in terms of virtue and human flourishing. In the following section, however, I argue that their conception of virtue is too dissimilar to ordinary skills for the practical reason of the former to be illuminated by that of the latter.
of what good agents (would) do in the circumstances, as opposed to what is dictated by rules or other such standards. And because these skills are so familiar and ordinary, modelling virtue on them dispels worries that it is impossible, or otherwise unsuitable, for moral agents to reliably determine what is called for in essentially the same way.
Trouble arises for agent-centered virtue ethics, however, when we consider the normative structure of familiar activities like these and what it implies about good practical reason in ordinary skilled domains. Consider cooking. No doubt, a good cook knows when a sunny-side-up egg has reached optimal consistency as well as what to do to achieve that end, even under suboptimal conditions. And we can grant that this practical knowledge cannot be exhaustively codified in a recipe, however long and detailed. Nevertheless, the cook's skill is assessed primarily by how conducive her actions are to producing a good sunny-side-up egg. If the whites are raw or the yolk is set, barring very unusual circumstances, she simply cannot count as skilled at cooking eggs. So, even granting that following instructions from recipe books is neither necessary nor sufficient for being a good cook, skillful cooking is nonetheless a matter of acting in a way that conduces to satisfying certain agent-independent standards.
Likewise, skillful building is conducive to constructing functional structures, like houses that withstand the elements. Excellence at sports and games is conducive to attaining certain goals in conformity with certain rules, like scoring runs in baseball and forcing checkmate in chess. Even skill at music and dance are largely a matter of acting in a way that conduces to satisfy such standards: staccatos must be crisp, and pirouettes must be fluid. Of course, it is possible to act skillfully in any of these domains without actually meeting the relevant success conditions. If the cook's stove malfunctions, a wildfire consumes the builder's unfinished house, the baseball or chess player is competing against an elite opponent, or the dancer or musician is injured, then they may not succeed despite their skill. Moreover, with enough luck, one might manage to succeed at such things despite having little skill. Even still, the most natural way to account for these possibilities is to say that skillful agency would satisfy the relevant success conditions, non-accidentally, under more-or-less typical conditions. In this way, agent-independent standards primarily determine what counts as skillful agency even in cases where skill proves neither necessary nor sufficient for success.
Modelling virtue on familiar skills like these, then, implies that virtuous agency, too, is primarily a matter of acting in a way that (non-accidentally) conduces to satisfying agentindependent standards. That is, act assessment, not agent assessment, is primary in the moral domain, just as it is in ordinary skilled domains. And if this is so, then the skill model of virtue is in tension with the primacy of agent assessment, setting up a dilemma for agent-centered virtue ethics. For this dilemma to hold, however, three further points must be clarified.
First, just as with the primacy of agent assessment, the primacy of act assessment can be understood as either a constitutive claim or a criterial claim. The constitutive claim is that the goodness (or badness) of the act makes the agent good (or bad); whereas the criterial claim is that the goodness (or badness) of the act is the best or only way to know whether the agent is good (or bad). For the dilemma to hold, the primacy of act assessment for skilled activities must be understood as a constitutive claim. This is because a merely criterial claim about the primacy of act assessment is compatible with the constitutive primacy of agent assessment, and thus with agent-centered virtue ethics.
To illustrate, proponents of agent-centered virtue ethics might grant that we need to know what fair acts are generally like in order to identify fair people, thus treating act assessment as primary in the criterial sense. But they would still insist that what makes an act fair is that a fair person does or would do it, thus maintaining that agent assessment is constitutively primary. So, for the skill model to pose a dilemma for agent-centered virtue ethics, act assessment must be understood to be constitutively primary for ordinary skills.
For instance, preparing food in a way that conduces to satisfying certain agent-independent standards (like set whites and runny yolks for sunny-side-up eggs) must make one good at cooking. It would be hard to deny that skill at cooking, and a wide range of other familiar skills, is indeed best understood in just this way. Riding a bicycle in a way that conduces to getting around safely and efficiently makes one good at cycling. Speaking French in a way that conduces to easy communication with Francophones makes one fluent in French. Playing chess in a way that conduces to forcing checkmate makes one good at chess. Certainly, experts can serve as role models for skills like these. And in some cases, they may provide the best or only evidence that a certain way of doing things is good, thus playing an indispensable criterial role for act assessment. But they surely do not determine, in the constitutive sense, what counts as good action in the first place. After all, skill involves more than just imitating experts. As Annas rightly notes: The learner needs to understand what in the role model to follow, what the point is of doing something this way rather than that, what is crucial to the teacher's way of doing things a particular way and what is not. (Annas 2011: 17) Indeed, the learner must grasp whatever makes a skillful performance skillful. And for a wide range of ordinary skills, it is clear that this is primarily a matter of acting in a way that (non-accidentally) conduces to satisfying agent-independent standards.
Second, it is worth reiterating that satisfying such standards does not entail following mechanical decision procedures. This is important because the inadequacy of such procedures in skilled domains is obvious, especially for highly embodied skills like sports, where agents must instantaneously react to complex and ever-changing circumstances. Explicit rule-following strategies simply will not do here. And even for less embodied activities like chess, skill is not simply a matter of applying rehearsed algorithms to select moves. Moreover, seasoned experts often determine what to do intuitively and automatically, and so they may not be able to explain why they acted as they did in a given situation (see Dreyfus & Dreyfus 1991).
Proponents of agent-centered virtue ethics might argue that this feature of ordinary skills parallels their claim that, because right action cannot be codified in rules, virtue is needed to determine what to do in particular situations. But, as noted in the previous section, act-centered ethics can also acknowledge that recognizing what is called for requires cultivated virtue, not just rote rule-following. What distinguishes these competing conceptions of virtue is instead the difference in normative structure: the primacy of agent assessment versus the primacy of act assessment. And while normative structure does have important implications for the intellectual structure of virtue (to be discussed shortly), the primacy of act assessment certainly does not imply that acting well is a matter of following mechanical decision procedures.
Third, it is not just that practitioners of ordinary skills are assessed by others according to agent-independent standards. Practitioners also assess their own actions primarily according to such standards. And this begins to explain how the normative structure of ordinary skills shapes their intellectual structure. Even when a seasoned expert cannot articulate exactly why she acted as she did in a given situation, she can certainly say why her action counts as skillful: "my whites are set, and my yolk is runny;" "my arpeggio was smooth and clear;" "checkmate!" This is so even when the action is not ultimately successful-"my roof would have shed water had the wildlife spared it;" "I would have hit a homerun if not for the elite center fielder;" and so forth. Agent-independent standards like these serve as the definitive success conditions in skilled domains. And the reason we can be sure that experts, and even novices, can readily cite them is that they determine what practitioners of all levels ultimately aim to achieve. 11 Furthermore, as success conditions, these normative standards provide the feedback necessary for skill development, defining how well an attempt goes, how effective a technique is, whether one strategy works better than another, and so on. Only by measuring their performance against such standards, and making adjustments accordingly, can practitioners hope to make non-accidental progress toward greater skill. Finally, even when skilled action is at its most intuitive and automatic, these standards still play the primary role in explaining how skilled agents learned to act as they do and thus why their actions count as skillful (see Fridland 2017). For instance, a cyclist learns how to balance better around turns by repeatedly executing such maneuvers, registering the results, and making adjustments to improve speed, stability, and other measures of successful performance. So even where the agent's precise reasons for action are opaque, including to herself, skillful performance is explicable primarily in terms of agent-independent standards of action.
To summarize, for practitioners at all levels of development, skillful agency must go beyond mere imitation of experts, focusing instead on the agent-independent standards of the given domain as the primary grounds for (self-)assessment. And grasping these standards in turn allows one to learn how to recognize and respond more effectively to the relevant features of one's circumstances as one gains skill in that domain. Only by understanding what counts as a good sunny-side-up egg can a cook learn to reason well about what tools, time, and temperature are best for making one, as well as what adjustments to make throughout the process to ensure, as far as possible, the desired results. This is what good practical reason involves for skillfully cooking eggs.
Nothing could be clearer or more ordinary. Similarly, good cycling involves reasoning well about shifting gears, applying brakes, leaning one's body, and regulating one's pedaling in order to get around safely and efficiently. Fluently speaking French involves reasoning well about grammar and word choice to express one's thoughts clearly, concisely, and in conformity with established conventions. 12 Many more examples are forthcoming. And while the practical reasoning of experts is certainly superior to that of merely competent practitioners in such domains, it is nevertheless informed by, and measured against, essentially the same standards. 13 Hence, observers as well as practitioners, experts and novices alike, all take the normative standards of the given domain to be agent-independent, and they assess themselves and others primarily according to them. 11 It is worth flagging here that, just as these normative standards serve as success conditions for skilled agency, human flourishing might be thought to serve as a success condition for virtuous agency. I will argue, however, in Section 4, that virtuous agents' flourishing is disanalogous to such standards in a way that makes such skills poor models for virtue, so conceived. 12 This appeal to convention might seem to ground normative standards for ordinary skills in agents after all. In a sense it does, but not in the sense required by agent-centered virtue ethics. It is one thing to say that standards for fluency depend upon the conventions of linguistic communities; it is quite another to say that such standards depend upon individual expert speakers. The former claim is more plausible, but only the latter makes fluency analogous to virtue as agent-centered virtue ethics conceive of it. I agree, then, with MacIntyre (1981) that virtue can be illuminated by practices like architecture, chess, and portraiture, all of which are assessed by conventional standards (though ones that are independent of any individual agent, including experts). But MacIntyre's dismissal of agent-independent standards like rights and utility as "moral fictions" indicates that his view of morality is agent-centered in a way that such skills are not. I thank an anonymous referee for pressing me to clarify this point. 13 Notably, as practitioners of ordinary skills get better, they often develop a better understanding of the relevant standards of action. And so while action is subject to assessment by the same standards for all practitioners, not all practitioners will grasp them in the same way. This point might seem to favor the agent-If this is what good practical reasoning is like for ordinary skilled activities, then the skill model promises a plausible and informative account of virtuous reasoning as a familiar capacity achievable, to a significant degree, by ordinary people. Thus, the skill model does exactly what its contemporary proponents wish it to do. But, understood this way, it also implies that act assessment is (constitutively) primary over agent assessment in the moral domain, just as it is in other familiar practical domains. Thus, what makes virtuous agency virtuous is that it conduces to satisfying agent-independent moral standards.
What's more, as with skillful reasoning, an informative account of virtuous reasoning is not forthcoming apart from such standards. Genuine virtue goes beyond mere imitation of moral exemplars. It requires grasping independent moral standards for oneself. And those standards serve as success conditions for action, determine what counts as good moral reasoning, inform how moral agents assess themselves, and provide the feedback necessary for moral improvement. Without such standards, in skilled domains and the moral domain alike, it would be unclear why one way of responding to a situation should count as better than another or how one might learn to respond to such situations better. And so, providing a plausible and informative account of virtuous reasoning via the skill model requires assuming the primacy of act assessment rather than agent assessment in moral life.

4
If this is right, then the dilemma holds: agent-centered virtue ethics must posit the (constitutive) primacy of agent assessment in order to maintain their distinctive theoretic appeal (first horn); but they must assume the (constitutive) primacy of act assessment in order to provide an informative account of good practical reasoning based on that of ordinary skills (second horn). And so, they must either give up the primacy of agent assessment or else provide a plausible account of good practical reason without appealing to ordinary skills. To avert this dilemma, proponents of agent-centered virtue ethics must argue that the normative structure of ordinary skills is not, in fact, relevantly different from that of virtue, as they conceive it. They could make this argument from either direction: either their conception of virtue is more like ordinary skills than I have allowed, or else ordinary skills are more like their conception of virtue than I have allowed.
Jason Swartwood employs the first strategy. In line with agent-centered virtue ethics, he understands practical wisdom (that is, the practical reason of virtue) as "an excellence that enables a person to make good choices about how to live…all-things-considered" (Swartwood 2013: 513). And he argues that this is "the same kind of epistemic achievement as expert decision-making skill" (512). He invokes psychologist Gary Klein's influential theory of expert decision-making, which "posits a two-stage process, starting with intuition, as decision makers recognize how they need to respond, followed by deliberate evaluation as they mentally simulate a possible response to see if it will work" (Klein 1998: 17). Decades of empirical research lend support to the hypothesis that this process enables experts in a wide range of practical domains to reliably make good decisions, even where their goals are vague or unstable. Expert firefighters, for instance, can evidently manage multiple Footnote 13 (continued) centered approach, on which specifying moral standards appropriately is a function of good practical reason. I argue in Section 4, however, that virtue, so understood, is importantly disanalogous to ordinary skills in this respect. discrete goals, like "ensuring firefighter safety, ensuring the safety of citizens, protecting property, and so on" as they decide which course of action is best, all things considered (Swartwood 2013: 525). If practical wisdom involves the same sort of process for making good all-things-considered decisions, as Swartwood suggests, then his strategy promises to avert the dilemma I have posed. His view of the practical wisdom of virtue makes it look more like the practical reason of (at least certain) skills than I have allowed, and so the primacy of agent assessment may be compatible with the skill model of virtue after all.
Swartwood's account of practical wisdom, however, is ultimately inadequate for the purposes of agent-centered virtue ethics. As Matt Stichter stresses: ...what is unique about practical wisdom is that it involves identifying which ends constitute living well, rather than what constitutes achieving those now fixed ends in specific situations. With the latter, you are trying to apply the conceptions of virtue you already have and make them more determinant to the specific circumstances you are acting in now...But with practical wisdom, it is a reflection that takes place during goal setting… (Stichter 2018: 132) In other words, practical wisdom involves not just means-ends reasoning but determining what matters in the first place, according to agent-centered virtue ethics. In skilled domains like firefighting, however, the ends are given, even if complex; and so expert decision-making is essentially a matter of determining how to achieve those ends in particular circumstances. Swartwood's strategy therefore implies that practical wisdom is likewise a matter of determining how to satisfy independently given, even if complex, moral standards rather than making more open-ended prior judgments about how one should live. His strategy therefore falls on the second horn of the dilemma: it uses the skill model to give a plausible and informative account of virtuous reasoning based on practical skills; but in doing so, it implicitly commits itself to the primacy of act assessment rather than agent assessment in skilled domains and the moral domain alike, thus undercutting the distinctive appeal of agent-centered virtue ethics. 14 Now, proponents of agent-centered virtue ethics might argue that the end of moral action is indeed given (and indeed complex): the good human life. And if so, then moral agents can presumably develop skill at living well in the same way they develop skill at other complex activities. But Stichter, for one, acknowledges: However, it is difficult to see how the feedback mechanism would work if wisdom is a skill in the sense of a singular all-things-considered judgment about how to act well morally. The specific problem is that the target of living well in that sense is very broad and vague, which will make it difficult to determine whether you are acting in such a way as to achieve success. (Stichter 2018: 133) Stichter therefore concludes that practical wisdom is importantly different from expert decision-making, even in complex domains like firefighting. He maintains, though, that virtue is a skill, albeit one that differs from other skills in that it involves ongoing reflection upon and refinement of one's conception of its end, the good life, and how one's particular actions are integral to that overarching project. 15 Stichter's view therefore falls on the first horn of the dilemma. It preserves the appeal of the agent-centered approach by making agent assessment primary: virtuous agents determine, through their practical wisdom, which ends constitute living well and thus which acts are right in their circumstances. But because it does so, it cannot elucidate virtuous reasoning on the model of ordinary skills. On such a view, a virtuous agent's conception of the good life is not only broad and vague, but also subject to ongoing reflection and revision, tailored to the individual, and encompassing all of practical life. The nature of the ends at which virtue aims, and its normative relation to those ends, is therefore radically different from the ends of ordinary skills. 16 As I have argued, such considerations have critical implications for the intelligibility of virtuous reasoning. The skill model provides a plausible and informative account of good practical reason precisely because relatively settled, agent-independent success conditions can be specified for ordinary skills like cooking, riding a bicycle, and speaking French, and plausibly for more specialized skills like firefighting as well. These success conditions provide measures for evaluating performance and, thereby, the feedback necessary for improvement. It is this normative structure that makes the intellectual structure of skill intelligible. And so, if virtue differs from ordinary skills in precisely this respect, then the reasoning involved in virtuous agency cannot be elucidated by analogy with them.
But proponents of agent-centered virtue ethics might push back at this point, using the second strategy noted above. Cheng-hung Tsai (2020), for one, argues that ordinary skills are more like virtue, as agent-centered virtue ethics conceive it, than my analysis allows. Specifically, he argues that just as one's conception of the good life is subject to reflection and revision, so are the ends of ordinary skills, since "one can deliberate about or choose among various specifications of the overarching ends" of such skills as well (Tsai 2020: 240). For example, the overarching end of swimming might be "having mobility and feeling comfortable in aquatic environments" (Ibid.), but a competitive swimmer will specify this end differently from someone who swims for therapeutic purposes. Tsai argues that this is so even for skills like chess, noting that beating a supercomputer would be an inappropriate goal even for a chessmaster. And so he concludes that for both virtue and a wide variety of ordinary skills, practical wisdom functions to specify the "best" end, "in the sense that the chosen or constructed specification is the most realistically, reproducibly, and/or challengingly achievable for the agent in question" (245). In other words, ordinary skills are best understood as having an agent-centered, not act-centered, normative structure. And if this is right, then the skill model of virtue is compatible with agent-centered virtue ethics.
The main problem with this strategy is that it implies an overly subjective account of virtue. Now, it is a notable feature of agent-centered virtue ethics, as Tsai notes, that a virtuous agent's conception of the good life will be somewhat specific to her, as will the acts she should perform in particular circumstances. For instance, it may be generous for one person to give $10 to a worthy cause, while for a much wealthier individual generosity might require $10,000. Nonetheless, any plausible conception of the good life is subject to certain constraints. One who is prone to selfish inclinations may not simply specify a conception of the good life in which charitable acts have no place. But this is exactly what Tsai's view suggests: if the goal of completing a single 25 meter lap can be appropriate for some swimmers, then a similarly low standard for living well, and thus for moral action, can be appropriate for some moral agents. That might be true if "appropriate" means "the best we can expect." But surely, such action cannot count as virtuous. 17 Moreover, even if we grant a highly subjective conception of virtue, this makes virtue importantly disanalogous to skill. In ordinary practical domains, succeeding at individually specified goals is neither necessary nor sufficient for exercising skill. A chessmaster who fails to beat a supercomputer can still show great skill by, say, lasting 40 moves. And a novice swimmer who succeeds at completing one lap does not thereby display much skill at swimming. In such cases, skill is assessed relative to standards that are independent of the ends specified by the individual. Thus, even if the success conditions for virtue are taken to be subjective, virtue is still importantly disanalogous to skill, pace Tsai. And so, the dilemma stands.

5
Despite the distinctive appeal of agent-centered virtue ethics, in fact because of it, their conception of virtue is simply too different from familiar, ordinary skills for the skill model to provide them with a plausible and informative account of good practical reasoning. Where to go from here? An insightful passage from Philippa Foot is suggestive on this point: To be sure, it matters a great deal, especially in personal relationships, how someone is rather than simply what he or she does…But given the horrors of the past century I think that today it would be especially strange not to see the 'what' of actions as even more important…It is no doubt of practical import to us to know what kind of a man can give the orders issued by Hitler, Stalin, Pol Pot, or Pinochet, and the personal evil of the legion of torturers now at their loathsome business in so many countries of the world-if only to know how we ourselves might come to act like that. But we do not need to know anything of that kind before branding the things that were and are still being done as utterly wicked. (Foot 2001: 113) This gesture to the primacy of act assessment is notable not only because of Foot's status as a pillar of contemporary virtue ethics but also because it comes at the end of her seminal monograph, Natural Goodness, a valiant attempt to ground moral value in the flourishing of the virtuous agent.
As I have noted throughout this essay, there are distinctive attractions of an approach to ethical theory on which "A conception of right conduct is grasped, as it were, from the inside out" (McDowell 1979: 331); which starts "with the virtues and vices rather than right or wrong acts" (Hursthouse 1995: 72); which "set[s] up the framework of the theory" this way and understands rightness in terms of virtue, rather than vice versa (Annas 1993: 9). But if this means giving up the resources of the skill model of virtue, then agent-centered virtue ethics must find some other way to defend their conception of good practical reason.
Alternatively, we might accept the primacy of act assessment and maintain the skill model of virtue. Foot's acknowledgement that the moral status of actions might be fully understood apart from a conception of virtue suggests that much of what agent-centered virtue ethics have to offer may be accommodated by an act-centered approach to ethical theory. As I have argued, the skill model of virtue can offer a plausible and informative account of good practical reasoning, based on that of ordinary skills, within such an approach. Moreover, I have argued that act-centered ethics do not entail an implausible conception of practical wisdom in terms of mechanical rule-following.
On the contrary, ordinary skills demonstrate how the primacy of act assessment gives rise to rich and sophisticated forms of practical reasoning. Understanding such skills on their own terms, rather than forcing them into the mold of a venerable, yet controversial, conception of moral virtue, would constitute a radical departure from much of the virtue ethics tradition. But it would nevertheless maintain focus on the kinds of questions that motivated the ancient architects of the skill model to develop it in the first place: What is practical excellence? What kind of reasoning does it involve? And what can ordinary skills teach us about good moral agency? Moreover, this strategy would renew the promise of a theory of moral virtue that is both attainable by, and suitable for, ordinary human beings.