PRACTICAL STRUCTURE AND MORAL SKILL

I argue that moral skill is limited and precarious. It is limited because global moral skill—the capacity for morally excellent behaviour within an ¨uber action domain, such as the domain of living, or of all-things-considered decisions, or the same kind of capacity applied across a superset of more speciﬁc action domains—is not to be found in humans. It is precarious because relatively local moral skill, while possible, is prone to misﬁre. My arguments depend upon the diversity of practical structures confronting human agents, the limitations of human skill learning and reason-sensitivity, and the failure of moral considerations to respect the social and institutional boundaries we develop to structure our practical lives.


I. INTRODUCTION
Many of us want to be good people-to have a morally good character, and to do the right thing much of the time. Few of us, I think, have it as an overriding goal to become as morally good as possible. But suppose that was your main aim: that you had as a goal a kind of moral analogue to the goal of many athletes, or artists. As the athlete and artist seek to develop skill(s) at their sport or art form, you seek to develop skill at moral action. How good could you expect yourself to become?
In this paper, I examine the structure and acquisition of moral skill, and what that implies about the prevalence, nature, and reliability of moral skill in human agents. The two primary theses I advance are that moral skill is limited in scope, and precarious. These theses break down along a distinction between global and local moral skill. This distinction depends upon a conception of the practical structure of much human action that I develop in this paper. To preview, I argue first that fully global moral skill, or anything approaching fully global moral skill, is impossible for human agents to develop. Second, I argue that relatively local moral skills are prone to misfire, hence precarious.
Note that these theses bear indirectly on the relationship between skill and moral virtue-a relationship that has received a good bit of recent attention (Annas 1995(Annas , 2011Daugherty 2020;Fridland 2017Fridland , 2021Fridland and Stichter 2020;Klein 2014;Stichter 2011Stichter , 2016Stichter , 2018Stichter , 2021aSwartwood 2013;Zagzebski 1996). Whether virtue is a skill, or is illuminatingly like a skill in some ways, is not something I consider here. My aim is more general, for one need not be invested in virtue theory to be interested in moral skill. And one might accept my thesis and think that virtue and skill should be closely linked, or not. Even so, the arguments below have indirect relevance for those who link virtue and skill. For such theorists, the arguments imply that like moral skill, virtue is both limited and precarious.
Section II offers some clarifications and qualifications regarding notions of skill, competence, and moral expertise. In Section III, I elucidate a mainstream view, shared by proponents of moral skill, regarding the nature of skill acquisition. In Section IV, I consider how we should think about the practical structures-the action domains and the agential modes-that are relevant to a notion of moral skill. In Section V, I argue that given these understandings of skill acquisition, and of the practical structure of action domains and agential modes, the acquisition of global moral skill is practically impossible for human beings. In Section VI, I grant that localized moral skills are possible, but I argue that such skills are precarious in much the same way as are skills at fast-reaction sport activities. In Section VII, I conclude, drawing a connection between my discussion of moral skill and Susan Wolf 's discussion of moral sainthood.
II. SKILL, COMPETENCE, MORAL EXPERTISE, ACTION DOMAINS, AGENTIAL MODES: CLARIFYING CENTRAL NOTIONS Most humans above a certain age have basic moral competence. I don't have a tight definition of moral competence, but my thought is that in everyday life, most people can regularly (although not flawlessly) discern the difference between the right and wrong (or obligatory, permissible and impermissible, or good and bad, or virtuous and vicious) action options available, and can regularly (although not exceptionlessly) do the right thing. This is a far cry from the thought that most humans above a certain age have moral skill. Compare the difference between those with non-moral skill and those with non-moral basic competence. Stanley Cup Champion and hockey all-star Ryan O'Reilly vs your mid-40s dad playing weekend shinny (that is, pick-up hockey at a local rink). The great Canadian portraitist Prudence Heward painting a portrait vs your uncle finishing his portraiture class. Some great philosopher (Ruth Garrett Millikan, say) vs an excellent undergraduate writing a response to a difficult philosophy essay. Supposing it exists, moral skill would require high levels of success at figuring out the best way to proceed-at discerning right and wrong (or at discriminating between shades of difference in the values presented by) action options, and at doing the right thing, even in difficult or unusual situations.
One might reject a strong distinction between moral skill and moral competence on the grounds that unlike specialized skills like hockey or painting that are difficult to come by, the natural developmental course of human agents imbues us with moral skill. Perhaps moral action is more like walking or talking, where the difference between basic competence and skill is slight or nil, and it makes sense to think that humans naturally develop skill. Thomas Reid may have had a view like this: From these self-evident first principles the whole system of moral conduct follows so easily, and with so little aid of reasoning, that every man of common understanding, who wishes to know his duty, may know it. The path of duty is a plain path, which the upright in heart can rarely mistake... There are some intricate cases in morals which admit of disputation; but these seldom occur in practice; and when they do, the learned disputant has no great advantage... (Reid 1788(Reid /2003 But moral behaviour does not seem so easy, and the natural competences that feed into moral action seem to fail us in too many situations to make this alternate view believable [for a similar point regarding expertise, see McGrath (2011)]. How many of us regularly fail morally when dealing with our kids, when dealing with our partner, when stressed, when tired? And how many of us have trouble figuring out what is right in situations that are unfamiliar or somewhat complicated? Speaking just for me and my friends, the answer is that all of us do.
Moral skill must also be distinguished from moral expertise. This latter notion has received more attention in moral philosophy, where people debate its existence, trustworthiness, and prevalence (Cholbi 2007;Cowley 2005;Driver 2006;McGrath 2011;Tiwald 2012;Watson and Guidry-Grimes 2018). For philosophers who discuss this issue, moral expertise has to do with the ability to conceptualize what is right and wrong in various situations, and to transmit this conceptualization in the form of advice. 1 Moral skill comes apart from the ability to give good advice, since some skilled behaviour contains key components that are not conceptualized by the skilled agent in the right way to function as good advice. 2 As I am using the term moral skill, then, it is closer to how Weinstein (1994) used the term performative expertise. Cholbi explains: Performative experts have the ability to 'get it right' within a particular domain, without being able to articulate or justify how the expert gets things right. The skilled marksman or musician displays performative expertise when she hits the target or executes a beautiful musical performance without being able to elaborate the steps by which she achieved these ends or even the criteria for excellence in meeting those ends. (Cholbi 2007: 325) Generally and roughly, then, skill is a capacity for excellent behaviour in some limited domain, where the standards for excellence hold within the domain in question. We are not yet speaking of moral skill, but we will get there.
Regarding the statement of skill in general, I take the notion of an action domain from Shepherd (2021a), who develops the idea as follows. An action domain is a cluster of action-types and goals that sets up common targets for skill learning. At the limit an action domain could be composed of a single action-type, such as running, baking, reading, or talking. Often an action domain is larger and more sophisticated. Agents become skilled at games (chess), at sports (tennis), at professions (journalism, philosophy, restaurant management) and activities within professions. So, a basketball player could be skilled at long distance shooting, a chef could be skilled at pastries, a professor could be skilled at teaching, or at administrative work, or at grant writing. An orthodontist could be skilled at dental surgery, or at diagnosis.
Various things-rules, institutions, shared goals or plans, certain human desires or needs-hold an action domain together, with variation depending upon the domain. What domains have in common, however, is that domains are structured in important ways around at least one ideal of success. This ideal can be self-imposed. It is possible to develop skill at some action domain that one invents (think of children inventing a new game). In the normal human case, however, these ideals are honed over time and handed down more or less officially. A key ideal in sports is to win. Commonly accepted ideals for those in the medical profession include the avoidance of harm and the promotion of benefit to patients. An ideal for many fields within academia revolves around knowledge-its creation, protection, and transmission. Of course in complex domains there may be multiple ideals of success, and these may sometimes come into conflict, creating practical difficulties.
Ideals plus circumstances common to a domain generate a kind of practical structure-a system of ends and means, weighted according to their relationship to the governing ideal(s).
Whatever the ideals in play, often they will imply, or will explicitly contain reference to, constraints of various sorts. These constraints can be built out of different elements. Consider the following four elements, common to more familiar action domains: [a] goal(s), [b] an ordering over the goals in terms of centrality-to-success, [c] restrictions on circumstance-types, [d] restrictions on behaviour-(and action) types. (Shepherd 2021a: 117) So, for example, in many sports, only certain surfaces, tools, and actiontypes are allowed. In most professions, institutional contexts restrict the kinds of actions one can perform and the kinds of circumstances in which one can perform them. Sometimes ideals of success in a domain imply or entail these additional constraints. Sometimes, they are added explicitly, as in the construction of a new board game. Sometimes, they are added implicitly, by processes of historical and social construction and accretion.
When I say that skill is a capacity for excellent or successful behaviour in some domain, then, one implication is that skill involves the capacity to navigate the practical structure proprietary to the domain in ways that enable high levels of success at meeting the central ideals of the domain. One might emphasize similar ideas with different conceptual resources. C. Thi Nguyen (2020) has recently written of the diverse practical structures set up by agential modes. As Nguyen has it, 'agential modes are tools for managing my attention, focus, and interests, and for controlling myself, in order to better bring my moment-to-moment practical reasoning and action into line with my genuine self 's interests' (Nguyen 2020: 85). An agential mode is, in a way, the subjective side of the practical normative structure we are interested in, while an action domain is the objective side. This need not imply a one-to-one correspondence between agential mode and action domain. Given the broadness and complexity of some action domains, we might think of an agential mode as a particular way that an agent fits into an action domain. There are different ways to be good at philosophy, or cooking, or basketball. In some cases, an individual agent will have only one mode prepared for some domain, and in some cases, this will be all they need. But in other cases an agent may have a few. One can imagine different modes being useful for an agent depending upon the 'space' of the domain the agent currently inhabits with switches between modes working to the agent's benefit.
In any case, the agential mode one deploys will bear a relationship to the normative structure of the action domain. As Nguyen observes, 'To enter an agential mode is to focus on a particular set of goals and on a particular set of abilities as the method for achieving those goals' (79). We might thus pull the ideas of an action domain and an agential mode together by saying that skill is a capacity for excellent or successful behaviour in some domain, facilitated by the appropriate use of an agential mode or modes. Saying this much places important emphasis on the normative structures proprietary to different skills-structures set up by action domains, and further tweaked by agential modes-without committing us to a particular account of skill. This is nice, because the recent literature on skill contains several competing proposals (e.g., Pavese 2016a,b; Shepherd 2021a; Stanley and Williamson 2017). These proposals differ regarding the essential nature of skill-whether it essentially involves certain kinds of knowledge, or relations to knowledge, or understanding, or behavioural control. But they all have to make room for the idea that skill involves excellent behaviour, by at least some standard of excellence. The way I prefer to understand it, the relevant standard(s) of excellence can be further elucidated by reference to the structure of the action domain in question, and also by reference to the structure of the agential modes upon which the agent relies.

III. SKILL ACQUISITION AND MORAL SKILL
As skill emerges within a domain, or regarding some agential mode, several things change within the agent. One idea I wish to emphasize in this section is that the agent's sensitivity to normative 3 reasons for action changes as skill emerges. This claim is in part inspired by a recent family of views of skill that emphasize not only processes of automatization and habituation, but also the continued relevance, and the increased efficiency, of processes variously described as conscious control, cognitive control, top-down attentional control, and practical reasoning (Christensen et al. 2016;Krakauer et al. 2019;Pacherie and Mylopoulos 2020;Shepherd 2015). And this claim is consistent with the fact that more than just an agent's sensitivity to reasons changes as skill emerges. In domains that require much bodily action, movements become more efficient. The body-the whole system of tissue, blood, and bone that drives movement along with the sub-systems that drive arousal levels, oxygen use, energy expenditure-changes in ways that tend to support better performance along a range of metrics. And these changes themselves change the structure of the reasons available to an agent. In the context of gymnastics (and probably much else), the considerations available to Simone Biles-the greatest gymnast of all time-are very different from those available to a past-prime philosopher like me. But let us focus on the part of skill that has to do with reason sensitivity.
I wish to remain as neutral as possible regarding the nature of normative reasons for action. According to Dancy (2018), reasons should be understood as three-place relations between some consideration (some state of affairs, or fact, say), an agent, and a mode of response (some way of acting, for example). What a reason does is to link the fact, agent, and mode of response via a relation of favouring. Some consideration favours, for some agent, a particular mode of response. Way (2017) argues that a reason is a fact that functions as a premise in good reasoning. Kearns and Star (2009) argue that a reason is a fact that functions as evidence that the agent ought to do something. All of these proposals share the idea that considerations indicate, in some way or other, that certain modes of response are normatively better in certain situations. It is this relatively theory-neutral idea I need.
If this is something like what a reason is, then sensitivity to reasons will somehow involve a facility with considerations that favour various modes of behaviour. In humans, this facility is complicated. One reason is that humans have many modes of behaviour available to them, and many ways these modes may be sequenced to satisfy goals, to produce success, or to produce failure (Israel et al. 1993). So the agent's facility with available reasons is going to be composed by a select package of perceptual, cognitive, and volitional capacities, organized in different ways depending upon the individual agent's constitution.
Now, this facility with reasons need not always be described in terms of the conceptualization of reasons as reasons. Shepherd (2021b), for example, distinguishes between psychological sensitivity and cognitive sensitivity, where only the latter involves explicit conceptualization.
An agent is psychologically sensitive to reasons (within some domain) if, across wide enough differences in circumstance, the agent's key psychological capacities -e.g., perception, emotion, attention -reliably track, in manners appropriate to the function of these capacities, the reasons that are available... Cognitive sensitivity is further nested within psychological sensitivity. An agent is cognitively sensitive to reasons (within some domain) if she tends to recognize or conceptualize available reasons as reasons -as considerations that recommend certain modes of behaviour -and if this conceptualization tends to drive and guide her reasoning and planning, as well as her behaviour. Now the agent need not have mastery with the concepts used to conceptualize some fact as a reason for action. But the agent does need some minimal level of competence with the relevant concepts, and with their relationship to forms of action, to qualify as able to conceptualize the reason for what it is. (Shepherd 2021b: 4) Some action domains, and some agential modes, may more frequently call for heavy-duty conceptualization and explicit reasoning and planning. Some action domains present reasons for action that call for rapid behavioural response, and some agential modes may be designed to handle exactly this. Agents may not need to conceptualize much in these domains or in these modes-they may do better if they train themselves to react immediately, rather than to think about how to react. In these cases, psychological sensitivity may be less informed by cognitive sensitivity, at least in the moment. But the agent may still qualify as reasons-sensitive, and she may still be able to cognitively access her reasons for action if she takes the time and effort to do so.
There is more one could say about reasons sensitivity in skill, of course, but I hope the basic idea is fairly clear by now. A further important question is this: How does the agent manage to acquire sophisticated forms of reason sensitivity?
Here it may help to turn to a group of theorists who are, in some sense, my opponents in this paper. This is the group that advances a line of thought according to which we can understand the virtues as moral skills (e.g., Annas 2011; Birondo 2021; Fridland 2017; Fridland and Stichter 2020; Stichter 2018). I do not comment on the connection with virtue here. I am more interested in how these theorists understand the development-the acquisition-of moral skill. For this, I turn to work by Matt Stichter (e.g., Stichter 2018, as well as work by Ellen Fridland (e.g., Fridland 2021;Fridland and Stichter 2020).
Their understanding of the acquisition of moral skills hews closely to the sciences of expertise and skill learning. So acquisition of moral skill is, in the final analysis, very similar to acquisition of non-moral skills. As Stichter has it, developing a skill involves 'internalizing standards about what counts as a good performance' (Stichter 2021a: 358). And this involves, in part, learning what the key goals and sub-goals in some domain are, and practicing how to achieve these goals repeatedly: 'This progressive mastering of subgoals requires "practice, practice, practice"' (359). Now, the sciences of expertise and skill learning emphasize not just practice, but a kind of deliberate practice (Ericsson 2008), as critical to the development of higher levels of skill. Stichter also emphasizes this idea: Deliberate practice requires having specific goals in mind for improvement, rather than the vaguer goal of 'getting better'. Deliberate practice is then a way of making progress by setting very specific goal intentions, and planning how to implement those goals in practice sessions. There need to be specific aspects of your performance that you go about planning how to improve, which then structures the kind of deliberate practice you engage in. (Stichter 2021a: 359) Deliberate practice involves structured cycles of action and feedback, the continued monitoring of progress, and the constant tweaking of one's abilities and habits. Fridland and Stichter (2020) emphasize the role of these elements of deliberate practice in generating key aspects of skill-situationally sensitive control and flexibility.
We emphasize continued deliberate practice in order to ensure that skills, and thus the structures that ground virtue, exhibit both control and flexibility. That is, we insist that skills are the sorts of things that we can easily manipulate and adjust appropriately in ways that are firmly connected to our goals. This aspect of skill is crucial for retaining and understanding the intelligence of skill in general and of virtue in particular. (6) Without situationally sensitive control and flexibility, a skill would not be the kind of thing that grounds excellent performance across a range of challenging situations. But Stichter, and Fridland and Stichter, repeatedly emphasize that this is what moral skills do. As they have it, [S]kills are distinct from many behaviours in their vicinity like habits or repeated stereotypical actions because skills result from deliberate practice which yields sustained improvements that are flexible in execution, and controlled in their systematic integration with an agent's goal states or intentions. This makes these sorts of actions intelligent and it also makes this model an appropriate one with which to think of virtue. Our claim is that the flexible and controlled actions that result from deliberate practice are different in critical ways from the products of other kinds of repetitive or rote behaviours. (Fridland and Stichter 2020: 7) This treatment of skill learning is mainstream, and accurate. Although there are important debates ongoing in the sciences that study skill acquisition-including debates about the scope of deliberate practice's impact-the account one finds in Stichter, and in Fridland, is pitched at a level of abstraction that largely avoids these debates. Almost everyone agrees that skill acquisition involves the progressive refinement and organization of cognitive structures, including one's mental models of relevant situations and one's own abilities to act, schematic knowledge of an action domain, as well as the progressive refinement and organization of one's abilities to put these structures to work via mental acts and mental actions involving attention, memory, imagination, inhibition, planning and selection of intentions, and so on (Christensen et al. 2016;Ericsson and Charness 1994;Watson 2020). As the agent practices, she not only acquires motoric capacities that make certain modes of response available and more fruitful, she learns about the structure of the domain so as to more readily conceptualize a reason for what it is. Learning structures the agent's capacities to find, recognize, assess, and in some ways reason with, the reasons that are present in a given situation. Learning is of course driven by the coordinated operation of a host of psychological systems-memory systems, obviously, but much more is involved-and is highly sensitive to task structure and to reward.
Further, almost everyone agrees that deliberate practice, which involves structured effort-feedback cycles, is an important part of skill learning (Ericsson 2008). Learning is driven by, reinforced by, reward and expectations of reward. As a part of this learning how to succeed, agents begin to cognitively map, both implicitly and explicitly, the relationships between several items: action options, action quality, common action obstacles, circumstance-types, action targets, successful outcomes, unsuccessful outcomes, and more [for recent work, see Behrens et al. (2018) and Mommenejad (2020)].
In summary, this model of skill acquisition-which I share with proponents of moral skill-is built upon the kinds of skills that expertise research is able to study in the lab and in the field. These are skills like dart throwing, free throw shooting, mountain biking, piloting a plane, or performing a specific type of surgery. On this model, then, moral skill emerges as a capacity similar to these other skills. 4 Now, if moral skill is like non-moral skill, then moral skill requires sensitivity to moral reasons for action, and the development of moral skill works in part by changing the structure of an agent's sensitivity to moral reasons for action.
An alternative is a view on which moral skill is a fundamentally different kind of thing from non-moral skill. Either it is not learned in the same way, or it depends upon a form of behaviour or an organization of action production capacities distinct from the form or organization that underwrites non-moral skills in whatever domain. I see no plausible consideration that indicates a different proprietarily moral form of behaviour in humans, nor any evidence that moral competence or skill come about by way of a different form of learning. Other theorists of moral skill seem to agree.
So we can say that moral skill is like non-moral skill, a capacity for excellent behaviour. 5 It depends, as does non-moral skill, on a near-optimal combination of background knowledge (and/or practical understanding), habituation, and reason sensitivity, for the domain in question. In the moral case, the knowledge, habituation, and reason-handling capacity will enable sensitivity to moral considerations. And the structure of the capacities will bear important similarities to the non-moral case.
But this raises a question about action domains. For non-moral skill, we have said, is a capacity for excellent behaviour as indexed to the standards of some particular domain. What domain-specific standards are relevant in the moral case?

IV. MORAL ACTION DOMAINS AND GLOBAL MORAL SKILL
Consider first the idea that the moral standards that apply to moral skill operate as a kind of extra action domain: just one more action domain alongside all of the others. You become skilled at squash and at punting and at chess, and you also become skilled at morality. This option is implausible. Reflection 4 See also Swartwood (2013), who argues that moral skill is practical wisdom at all-thingsconsidered decisions, and is acquired in the manner of all skills: 'wisdom is developed in the same way as expert skill: through deliberate practice that gives a person feedback on the quality of their decisions' (2013: 523). 5 I am not here committing to anything substantive about morality, except that morally excellent behaviour is possible. I am not committing to any normative theory, nor even to the idea that there is a correct normative theory. I am not defining morally successful behaviour in terms or the right or the good or the virtuous. And I am not yet saying anything about the domain(s) at issue for morally skilled behaviour. I am only trying to make plain the relationship between moral skill and reason-sensitivity. on the way that moral reasons arise within various action contexts suggests that though moral considerations are infused in different ways and to different degrees within different action domains, moral situations arise within any action domain.
Consider, then, the possibility that morality is anüber action domain. For moral action takes place-moral reasons arise-amongst the many action domains that structure our social and practical lives. We find morally loaded situations in the dentist's office, in the hallway of the university, on the rugby field, at the barstool, at the family reunion.
Even if morality is a kind ofüber domain, it seems that the practical structures proprietary to each domain influence the types of moral considerations that arise, and the frequencies with which they do so. Some domains are primarily non-moral. Domains set up by games, or by some arts and crafts, rarely offer moralized situations qua domain.
Other domains are moralized in specific ways, such that skill in that domain requires reason-sensitivity regarding certain specific moral considerations. A specific set of moral considerations is arguably embedded into educational professions. There is a kind of pastoral responsibility professors bear to students, but it is scaffolded by the aims of the university, and by features of the professor-student relationship. A specific but distinct set of moral considerations is embedded into public service professions, or in medicine.
Nguyen emphasizes a similar idea, thinking of the issue in terms of agential modes: Often, doing the right thing involves finding an appropriate agential mode. When talking with a student during office hours, I might realize that our conversation isn't just about the details of a paper, but that they are actually profoundly distressed and emotionally overwhelmed. I need to switch from philosophy-teacher mode -focused on teaching rigorous arguments and writing clarity using the tools of argument analysis -to a more therapeutic mode, where I aim at finding and easing their emotional distress using various empathetic abilities. (Nguyen 2020: 79) This is a useful way to think of a situation we are commonly in. In some action domains, we may need to switch agential modes on-the-fly in response to overriding moral considerations. Some action domains may call for this kind of switch more often than others. Some action domains are prone to moral situations-situations when moral reasons are much more relevant than practical, non-moral reasons-but unevenly. Given duties to of care to young children, domains surrounding childcare are plausibly like this. And clerical work in organized religions often places one into contact with others in ways rich with moral implications.
The idea here, then, is that a moral skill (or a set of agential modes that facilitates excellent moral behaviour) prevalent in domain A may not much resemble a moral skill prevalent in domain B in terms of the patterns of reason-sensitivity agents display. By way of clarification, consider recent work by Daugherty (2020), who argues that virtue can be fruitfully analogized to skill, provided the skill in question is restricted to skill in certain roles.
Contemporary Western life and society are perhaps less explicitly structured around our roles than they once were. Nonetheless, I take it that we do still occupy roles and that we do still have a basic grasp on the notion of a role. We know what it is to be a parent, a citizen, or a carpenter, for instance, also a cobbler, doctor, or tennis player. And as at least some of these roles properly involve skill, we should also have a basic grasp on the notion of a skill role. (Daugherty 2020: 85) This notion of a role is obviously traversing a space similar to the notions of an action domain and an agential mode. For Daugherty, a skill role is restricted in important ways. Within a role, agents 'serve some function in a practice' (86). For example, doctors treat patients. Cobblers make shoes. These practices have a kind of structure that makes demands of the agents who fill the roles: 'since a doctor's [role] is to see and treat patients, doctors ought to see and treat patients; and since a cobbler's is to make shoes, cobblers ought to make shoes' (86). To be a good skill-role occupant is, at least in part, to meet these demands-if you like, to meet certain ideals that lend practical normative structure to the role.
For Daugherty, however, there is a difference between common skill-roles and moral skill. For virtuous action exists in a kind ofüber-domain structured by the practice of living well.
[G]ood skill-role occupancy is analogous to virtue. As a human being's activity is living, the virtuous human's distinctive skill would be "skill at living"-the know-how they possess, knowledge how to live. (99) Let us call the kind of moral skill Daugherty is aiming at-and the kind many virtues-as-skills theorists seem to aim at as well-global moral skill. We could understand global moral skill as moral skill regarding anüber action domain, such as the domain of living. Or we could understand global moral skill as a superset of action domains-something like the superset of those domains relevant to the distribution of moral reasons for action across the space of practical possibility for an adult human life. There will be ways of making this notion more precise, of course. One might carve out restrictions regarding the distribution of moral reasons, or regarding what kind of adult human life might matter for some articulation of moral skill. One might also wish to specify the skill in question as skill at utilizing, and transitioning between, the wide range of agential modes relevant to the distribution of moral reasons across the space of possibility. And one might emphasize, as seems plausible, that moral skill conceptualized in this way comes in degrees-so one's moral skill may be more global if, for example, it covers a bigger superset of morally relevant action domains, or if it covers a broader range of situations that arise within theüber domain. I do not think we need to be any more precise than we have been, for present purposes.
I turn to the argument that global moral skill is practically impossible for human agents.

V. GLOBAL MORAL SKILL IS PRACTICALLY IMPOSSIBLE
Global moral skill is limited in that anything approaching fully global moral skill is practically impossible for human agents to develop. Humans lack the psychological resources to develop moral skill across the collection of (all or most of) the action domains that structure human practical life. This is because, as the discussion above regarding the wide variety of action domains makes clear, there are simply too many morally relevant action domains to permit human mastery across such a broad group. There is too much variation in practical structure across differences of domain; there is too much variation in practical structure across the agential modes one might fine-tune with moral excellence in mind.
Consider the following facts about human skill learning. First, skill learning takes time. In complex domains, it takes a great deal of time. Second, skill learning is domain specific. Many domains require the development of specialized knowledge-both non-moral and moral-as well as specialized coordination of perceptual, motivational and cognitive capacities. This specialized knowledge and psychological coordination does not cleanly transfer to dissimilar domains. The skilled hockey player starts nearly from scratch when approaching chess. And the morally skilled professor starts nearly from scratch when approaching the battlefield. Although learning from one domain can transfer to-and facilitate performance in-another domain to some degree (Abernathy et al. 2005), this depends upon the similarity of domains along relevant parameters. The more dissimilar an action domain from the domains of an agent's skill, the less like an expert the agent will perform. And there are transfer costs even for fairly similar domains (Rosalie and Müller 2012;Schmidt and Young 1987). Third, skill learning requires a fairly rigorous and stable structure of immediate feedback in terms of costs and rewards. We noted earlier that reward and punishment are largely what drives learning. So one must know what constitutes reward and punishment, and this should be compelling. This is, in part, why deliberate practice is important. In an important review, for example, Ericsson notes that experience alone in some domain is not a great predictor of skill development. A much better predictor is time spent in deliberate practice-that is, conditions in which agents are '1) given a task with a well-defined goal, 2) motivated to improve, 3) provided with feedback, and 4) provided with ample opportunities for repetition and gradual refinements of their performance' (Ericsson 2008: 991). For many action domains, we are able to provide something like this. But we rarely have available rigorous and stable feedback regarding specifically moral ends and values as they arise within a given domain (Swartwood 2013).
A further difficulty is this. We learn more or less constantly. And moral action, along with some degree of feedback regarding that action, is ubiquitous. The basketball player does not easily pick up bad basketball habits when she leaves the court and returns to the classroom. But the moral agent has no comparable court of morality. She leaves church, but she does not leave the moral arena. And the feedback she receives is not stable across the contexts and domains of her practical life. Dan Jacobsen makes something like this point in a complaint against the analogization of virtue to skill.
Habituation into virtue works because emotional rewards and sanctions gradually alter a person's affective responses and motivational tendencies, in ways that can correct them. Yet people regret decisions that turn out badly according to their own criteria and triumph in success judged by their own lights -not to concordance with some independent "space of reasons." Granted, some forms of feedback can arise from other sources, most notably the culture in which we live. Shame and guilt in response to the contempt or anger of others, for instance, along with pride in response to positive social recognition, help condition our responses so as to accord better with the expectations of peers and authorities. But social feedback cannot inculcate such robust practical wisdom either, since the socially accepted consideration may not be the truly salient one. (Jacobson 2005: 400) In a recent symposium on Stichter's (2018) book, Mark Alfano (2021) puts a very similar criticism to Stichter. Stichter's reply is interesting. First, he admits that acquisition of moral skill is difficult. Second, he claims that the focus of skill acquisition, in the moral case, should be competence, not skill: that it should be 'trying to get people above a minimal threshold of virtue' (Stichter 2021b: 612). Third, he claims that competence will be easier to acquire, in part because 'I suspect that there's going to be more equivocal and accurate feedback on immoral actions' such that 'virtue acquisition is going to be easier at the initial levels of development' (613). This response seems right as far as it goes, but it fails to reckon with the force of the criticism. If competence is the goal, why talk of skill? And if skill is developed via structured regimes of deliberate practice and clear feedback, why think moral skill is possible? Development of levels of skill that leave mere competence in the background require not just punishment, but reward-feedback that clearly marks out subtle differences between success and failure and that allow agents to progressively refine the structure of their abilities and habits. It is doubtful that fine-grained feedback such as this is available in daily situations.
So we have reason to doubt the possibility of fully-or even fairly-global moral skill for humans. I do not deny that such skill is available in principle to some psychologically advanced agent, embedded in some morally advanced society able to provide the right sort of practice regime. I only doubt it is available to humans. Moral skill for humans, it seems, is limited to particular (clusters of) action domains, or to particular (clusters of) agential modes. Moral skill for humans is only (relatively) local.

VI. LOCALIZED MORAL SKILL IS PRECARIOUS
Here the distinction between global and local moral skill may be of some help to the virtue ethicist. For even if anything approaching global moral skill is practically impossible, localized moral skills may be available. In specific action domains, the structure of situations may be regimented enough to permit better understanding of common moral reasons, and to allow for clear feedback on behaviour.
For present purposes, I am going to assume that localized moral skill is possible. It seems to me a substantive task to say how such skill will look in particular situations-what kinds of agential modes are morally beneficial for, e.g., the family doctor, the philosophy teacher, the union organizer, the children's softball coach, and so on. I am going to leave that to one side. All I aim to do in this section is offer a characterization of localized moral skill as precarious.
Consider, as a way into my line of thinking here, this example. The person, who spends his life as a ski patroller in winter and an adventure rafting guide in the summer, will become familiar with certain sorts of situations-both non-moral and moral. These might be situations calling for patience with tourists, or courage in the midst of physical danger, or empathy and calm when dealing with someone who is badly injured. We might expect that a morally sensitive ski patroller has available various agential modes that would also be of use in adventure rafting, and perhaps also in unrelated action domains, like waiting tables at a busy restaurant. But compare the morally excellent ski patroller with the morally excellent professor. The professor has a facility at navigating power dynamics between students, colleagues, and administrators, and is able to structure interactions with students to promote the somewhat moralized aims of education. She learns how to mentor graduate students and postdoctoral fellows so that they remain confident, upbeat, motivated, engaged in research. She is also empathetic, charitable, and honest with colleaguesand much more besides, depending on the needs of the situation-able to promote the kind of relationships that lead to a well-functioning department. (This person sounds too good to be true, I know.) I would not claim that moral considerations have no generality across domains. But it does seem plausible that the morally excellent ski patroller would be out of his depth-both morally and non-morally-if switched with the morally excellent professor, and vice versa. This is because we should predict relative differences in the typicality of moral considerations depending upon the domain. Although the structure of moral ends is stable enough to offer guidance in most situations, we should predict that moral excellence does not easily transfer.
A further difficulty is that, as Daugherty notes, the demands that a particular skill role places upon the agent depend upon the structure of the role. The interaction with moral considerations complicates matters.
[C]learly not all role demands are moral demands... the demand on a doctor to do such-and-such for patient x with ailment y is... a demand of medicine; and the demand on a cobbler to make shoes and do so in way z is a demand of cobblery. Some of these demands may then also be characterized as being of moral concern, but as role occupants are subject to their specific role demands only in virtue of occupying those positions in the practice, the practice will be basic in this respect. (Daugherty 2020: 87) Skill is developed by repeated interaction with the structure of costs, rewards, success, failures, and reasons common to an action domain. But in everyday life, even in the context of some specific action domain, or skill-role, or some specific agential mode, moral considerations are not restricted to the specific structure of that domain or role or mode. We often lose control over the action contexts that confront us. We often find ourselves in contexts that call for moral skill that applies across more than one type of domain. And it is often difficult to determine whether some action context calls for a mode of behaviour that is familiar, or a mode of behaviour that is beyond the boundaries of our skill. Nguyen notes that while 'cognitively limited beings like us usually approach the world one agential mode at a time' (Nguyen 2020: 79), often in the 'freeform' natural world we have to understand which mode to deploy: 'Often, doing the right thing involves finding an appropriate agential mode' (79). The simple point I am making here is that we often lack the time and the wherewithal to practice shifting between modes in ways that could qualify a localized moral skill as highly robust. The moral considerations that may arise in any given situation may transport one beyond the skills one has learned in the context of any particular role, or mode, or domain. Moral reasons fail to respect the social and institutional boundaries we develop to structure our practical lives. Our difficulties are compounded by the lack of rigorous feedback in many domains, and also by our own theoretical and personal uncertainties about the structure of morality.
For these reasons, the kinds of moral skill we are able to develop strike me as analogous to skills at rapid-reaction sports. There is a lot of work on such skills in sport psychology, and it is familiar to any fan of these activities that even the best agents fail frequently, simply because they are in such difficult conditions. When returning serve in tennis, or hitting in baseball or cricket, the agent faces conditions of limited information, limited time, and limited ability. Highly skilled agents work to move their success rates up slightly, while knowing that their skill is precarious.
Localized moral skill seems like this to me. It is genuine, but it is prone to misfire. So, for example, the morally skilled parent may enjoy some success. They may read books about parenting, develop by trial and error a standard set of reactions, interventions, and expectations, and agential modes. But their skill may nonetheless frequently give out. I, for example, possess the goal of being a morally skilled parent. I believe that I work as hard and as smart as I can to become such a thing. And yet I do not always know how to shift between the relevant agential modes-friend, disciplinarian, wise old sage, patient older brother-type, defender, laid-back freedom-giver-that may be required. Nor do I always feel like I have put in enough work at whatever agential mode would be useful in some new situation. When the child enters a new developmental phase, or makes a new friend, or confronts a novel challenge at school, or when the parent is unusually tired or stressed, the skill may give out. Whether I am morally skilled as a parent, I genuinely do not know. But even if I am, it is a precarious sort of skill. I often swing and miss, so to speak.

VII. CONCLUSION
I have argued that moral skill-the capacity for morally excellent (or praiseworthy) behaviour in some domain-is both limited and precarious. It is limited because global moral skill-the capacity for morally excellent behaviour within anüber action domain, such as the domain of living, or the same kind of capacity applied across a superset of action domains-is not to be found in humans. It is precarious because relatively local moral skill, while possible, is prone to misfire. My arguments depend upon the limitations of human skill learning and reason-sensitivity, the diversity of practical structures confronting human agents, and the failure of moral considerations to respect the social and institutional boundaries we develop to structure our practical lives. I wish to close by considering a resonance between this view of moral skill and Susan Wolf 's views on the moral saint-that is, 'a person whose every action is as morally good as possible' (Wolf 1982: 419). One of Wolf 's sub-arguments is that, for practical reasons, the moral saint will be unable to develop all of the moral virtues as well as any significant number of interesting non-moral virtues, skills, or qualities.
[T]he moral virtues, given that they are, by hypothesis, all present in the same individual, and to an extreme degree, are apt to crowd out 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. (421) Wolf finds this objectionable-a life without Victorian novels, or gourmet cooking, or a love of fashion, or whatever, 'may seem to be a life strangely barren' (421). Ultimately, she thinks, the moral saint will be either 'too good for his own good,' or at least 'too good for his own well-being' (421).
I do not disagree. But notice that, while Wolf assumes that a moral saint with very many non-moral skills or qualities is practically impossible, she assumes that a moral saint-someone who manages with each action to approach the best possible moral outcome-is practically possible. What I have said about moral skill undermines this assumption. 6 If (as I have argued) global moral skill is a practical impossibility for human agents, then a moral sainthood that applies across the board is impossible as well. For one is a moral saint-indeed, one approaches moral sainthood-only if one possesses moral skill.
Local moral sainthood may be possible. Will the localized moral saint be too good for their own good? Will they lead a life strangely barren? This will depend upon the kinds of agential modes that support moral skill in various action domains. I have put that intriguing issue to the side. But consider the possibility that the capacities that structure moral skill in some localized domain depend to some degree upon the development of non-moral skill in that domain. One cannot develop moral skill as a ski patroller without developing significant skill at skiing, and at managing dangerous situations generally. Perhaps one cannot be a morally skilled philosophy professor without some aptitude for philosophy. I am not certain that this hypothesis that the practical and moral structure of various action domains (or agential modes) intermingle is true across the board. But if one cannot develop moral capacities to a high degree of sophistication without developing concomitant non-moral skills, then it will be difficult to become a (local) moral saint without developing non-moral qualities that add richness to life. 7 The practical difficulty of becoming a moral saint, and the relationship of moral capacities to non-moral skills, might suggest that the person who wishes to even approximate moral sainthood must take deep dives into the (non-moral) world as it is, and must seek to understand human beings (and themselves) as they really are. 8 If so, they will not be able to become bland in the way that worries Wolf without undermining their capacity to perform high-quality moral actions. 9