1. Introduction: reason and will

We all know Hume says,

whereas Kant says,

They are commonly taken to be in a dispute about the role of reason in practice: whether this consists merely in determining means suitable to ends given from elsewhere, or also, and more fundamentally, in determining what ends to be pursued. How do we adjudicate such a dispute? When one party claims that rational reflection can and should play only an instrumental role in deliberation, while another claims that it can and should determine what ends are to be pursued, where do we look to figure out who is right?

An influential body of work in contemporary ethics and practical reason proposes that the dispute can be resolved by reflecting the nature of agency. The idea is that we might establish the authority of certain practical requirements – perhaps only conditional requirements between ends and means, or perhaps unconditional requirements on our ends themselves – by deriving them from the the very idea of agency. Seen through this lens, the dispute between Hume and Kant is about whether an agent, considered simply as such, is subject only to requirements to take means conducive to its ends, or whether it must also be subject to further, substantive requirements on the ends it pursues. The dispute, in other words, concerns what standards of correctness in practical thought, or measures of acting well, belong internally or constitutively to the power to act.

On one widespread conception, constitutivism is the project of defending a thesis of precisely this sort: a thesis that basic practical requirements are grounded in the nature of action.Footnote ³ Constitutivists typically argue, in addition, that fundamental practical principles can only be validated in this way – that an agent could shrug off without inconsistency any requirement that could not be traced back to the very nature of agency. Even in the absence of such an argument, this way of understanding the authority of a standard of action has obvious appeal. If it could be shown that any agent is, by nature, subject to certain requirements, this would elegantly clarify why we, who know ourselves to act intentionally, must (in some sense) know ourselves to be bound by them. And if there were such a proof, it would provide a particularly attractive sort of vindication: it would show them to be objective, while at the same time illuminating why they are such as to immediately engage the will of each (perhaps not always, but when we deliberate without interference). It would thus avoid a familiar dilemma: the apparent need to choose between views which vindicate the objectivity of standards of action in a way that leaves a residual mystery about how they might make a dent on the will, and views which explain the immediate bearing of normative thought to move us to act on motivation, but in a way that deprives them of objectivity. Even if some philosophers have given up one or the other, they have done so under the pressure of the difficulty. The natural first response is to seek an account that shows the incompatibility to be merely apparent.

Can the concept of action bear the theoretical weight constitutivists place on it? One ground for skepticism is that different constitutivists come to such strikingly different conclusions about which requirements are built into the structure of agency. Humeans insist that only hypothetical requirements to take appropriate means to one's actual ends, whatever they may be, can be part of this structure.Footnote ⁴ Kantians maintain, on the other hand, that being subject to categorical requirements on ends is a condition of the possibility of being an agent at all.Footnote ⁵ There is reason to worry that intuitions about the requirements contained in the concept of agency simply reflect substantive convictions about the normative requirements to which we are subject, and cannot supply an independent point of leverage on the latter question. Moreover, even if it were shown that, on some conception of agency, being an agent involves being bound by certain principles, it might still be asked what reason there is to be an agent in the relevant sense. A number of critics have argued that, in the absence of an answer, a putative derivation of normative requirements from the concept of agency cannot show them to be unconditionally binding. And, the critics add, answering this question requires appealing to a source of normativity other than, or external to, agency itself, and so the constitutivist program simply cannot succeed.Footnote ⁶

2. Three assumptions about constitutivism

The aim of the present essay is to raise some questions about a set of assumptions accepted by all the parties to this dispute – Kantians, Humeans and their rationalist critics. They assume that an adequate constitutivist account must meet the following conditions:

These assumptions are closely related. Someone who accepts Conceptualism must accept both Formalism and Absolutism, and while a commitment to Formalism or Absolutism may not strictly imply the others, it certainly makes them natural to think. In any case, my project here is not to address subtle issues about the logical relations between these commitments, but to ask whether any should be accepted.

I raise this question from a standpoint deeply sympathetic to the spirit of constitutivism. I think it is right to see the debate between Hume and Kant as a debate about the kinds of practical requirement to which we are subject in virtue of our capacity to act for reasons. And I agree that understanding the basis of practical normativity in this way would provide a particularly elegant and attractive way of addressing various basic problems in ethics and meta-ethics. My aim is to show that the project of constitutivism can be understood more broadly than its contemporary advocates and opponents typically assume, a way that retains its deepest attractions, while avoiding various difficulties that beset its most prominent contemporary incarnations. If I am right, there is room for a pluralist, non-formalist constitutivism that that does not seek to ground normative requirements on action in the sheer concept of rational agency, but that still presents appealing and distinctive responses to fundamental questions in ethical theory. Moreover, seeing this possibility will shed a revealing light on the debate between Hume and Kant, and the kind of resolution it might admit.

My conviction that such a view is possible grows out of reflection on Aristotelian ethics. Contemporary discussions of constitutivism seldom mention Aristotle and the kind of naturalism he espouses, but he surely deserves to count as a constitutivist in some important sense.Footnote ⁷ His view is, famously, that the fundamental norms governing how to lead our lives are grounded in the fact that we have by nature a certain ergon or characterisitic activity, and that the structure of the human soul implies a certain appropriate division of labor in how do what it is our to do. Thus, for Aristotle, the notion of the nature of a human being plays a crucial role in explaining practical normativity, or again the good in action: the fact that we, human beings, act well in doing in certain ways is grounded, ultimately, in what we essentially are, where this is characterised in terms of what we essentially do, our characteristic activity.Footnote ⁸ In this sense, Aristotle holds that the norms to which we are subject are constitutive norms. It seems, however, that his conception of our constitution, of what we are, includes more than the mere fact that we are capable of acting for reasons: he locates the ground of the practical norms to which we are subject, not in our nature as agents, but in our nature as human beings, a certain kind of material being. Moreover, he displays a striking lack of concern with the task of showing that the relevant norms apply to all possible agents, still less with the task of deriving the relevant norms from some abstract or formal characterization of our nature. He famously says that he aims to speak only to persons who have received a sort of upbringing that already equips them to see the point of certain ways of acting, and he seems to allow for the possibility that different life-forms might imply different lists of ethical virtues. I think these features of Aristotle's project are not simply independent commitments, but expressions of a single underlying idea – hylomorphism, to give it a name. A central aim in what follows is to bring out the bearing of this idea on the contemporary debate about constitutivism.

3. Absolutist constitutivism: Kantians and anti-Kantians

4. The possibility of pluralism

To assume that a vindication of Hume can be inferred from Dreier's premises is, in effect, to assume that if only the instrumental principle is necessitated by the very concept of a capacity to reason practically, then the instrumental principle is the only principle that can belong necessarily to a capacity for practical reason. It is to assume that if no further requirement of practical rationality is conceptually necessary, then no further requirement of practical rationality is actually possible. This is to overlook the possibility of a pluralist constitutivism, one which acknowledges that a Humean form of practical reason is conceptually possible, and thus that no unconditional practical requirements can be derived from an analysis of the very concept of a rational agent, but which maintains that other, richer forms of agency are also possible.

Dreier does not give an argument to rule out this possibility: he simply assumes that the debate over constitutivism is a debate about whether there are practical requirements to which all conceivable rational agents are subject. And though I cannot argue the point here, I think this holds true also of other prominent contemporary anti-Kantians: they pass from arguments for the claim that Humean rational agents are conceptually possible to conclusions about what practical reason demands of us, by way of an uncritical assumption that any constitutivist vindication of practical requirements must ground those requirements in the very idea of a rational agent.Footnote ¹⁸ This absolutist assumption, shared by contemporary Kantians and anti-Kantians, gives their dispute the appearance of a conflict in which only one side can be correct. But once we recognize the possibility of a pluralist constitutivism, we can see that each side might be right about something significant. The anti-Kantians might be right that the very concept of a capacity to act for reasons does not imply any further requirement beyond the instrumental principle, while the Kantians might be right that our actual capacity to act for reasons takes a form that implies the existence of unconditional practical norms. So long as neither side frames its point in absolutist terms, their views are consistent.

5. The problem of inescapability

If the discussion of the last section is sound, a pluralist position – one that allows for the possibility of multiple forms of agency, each constitutively subject to its own characteristic requirements of practical rationality – is a dialectically attractive option that is mostly overlooked in contemporary debates over constitutivism. This calls for explanation: if this sort of position is available and attractive, why is it not considered?

To see a reason for this neglect, it will help to consider an influential criticism of the constitutivist program. This criticism brings out that, on some ways of understanding the thesis that being subject to certain norms is ‘constitutive of being an agent’, this thesis proves too flimsy to support the metaethical results that constitutivists hope to derive from it. It can seem that the only way to avoid this sort of flimsiness is to embrace Absolutism, Conceptualism, and Formalism. I conjecture that it is some – perhaps tacit – recognition of this difficulty that prevents the pluralist alternative from coming into view. But in fact, I will argue, we can avoid the difficulty without embracing these commitments.

The kind of objection to constitutivism I have in mind is exemplified by David Enoch's much discussed essay ‘Agency, Shmagency’. He begins by noting that a main attraction of constitutivism, as it is standardly presented, is that it would clarify why certain normative requirements on action are rationally authoritative – why a rational agent cannot coherently ask, concerning such norms, ‘Why should I care?’ For, constitutivists hold, since the question ‘Why should I care?’ is asked by an agent considering how to act, and since the relevant norms are implied in the very idea of agency, there can be no genuine question about whether, in the relevant sense of ‘should’, these norms should be followed. But, Enoch objects, this sort of account of the authority of certain norms simply takes it for granted that one should strive to be what ‘an agent’ ought to be. If being an agent requires conformity to certain norms to which many people do not actually seek to conform, then it seems coherent to ask ‘Why should I be an agent?’ But if this question is coherent, then any norms implied by the idea of agency appear to apply to a given subject only conditionally, insofar as she has reason to be an agent. In the absence of such a reason, it is rationally permissible for her to say ‘Agency, Shmagency!’ And even if there is such a reason, it presumably cannot itself be grounded in norms implied by the very idea of agency, since it is supposed to ground the applicability of those very norms. So in either case, the constitutivist project of grounding the authority of certain norms in the nature of agency fails.

Now, it is not clear that Enoch's question, ‘Why should I be an agent?’, is genuinely coherent. Constitutivists generally reply that being an agent is not an attribute that a subject can coherently choose to have or not to have. I am an agent, not by choice, but simply by virtue of my nature as a free, self-determining being. Any choice I make will inevitably be the choice of an agent. So, arguably, the question ‘Why should I be an agent?’ presupposes the possibility of a standpoint that no deliberating subject can actually occupy. If this is right, then any norms implied by the very idea of agency are not just conditionally binding; they are binding on any subject who can confront the question what to do, and the constitutivist project is vindicated.

This reply, however, may seem to be available only to a constitutivist who relies in his argument on the most abstract and minimal conception of agency. For it may seem that being an agent is an inescapable attribute of a subject who confronts the question what to do only if ‘being an agent’ means nothing more than: being such as to be capable of confronting the question what to do. After all, given any richer conception of agency, it is not a necessary truth that a subject who can act for reasons must be ‘an agent’ in this sense. So given any richer conception of agency, Enoch's question may seem to regain its grip: why should one strive to be ‘an agent’ in this richer sense, if it is possible to deliberate, choose, and act without being such a thing? But if Enoch's question gets a grip, then Enoch's dilemma for constitutivism applies. So it appears that a principled constitutivism must aim to ground whatever norms it seeks to vindicate in the very concept of agency, a concept that applies to any conceivable subject capable of acting for reasons. And then a pluralist constitutivism is obviously unacceptable: for if a certain form of agency is supposed to be, not the one-and-only form required in any agent capable of acting for reasons, but one of a plurality of possible forms, then the normative requirements implied by that form will be open to Enoch's challenge, and so the relevant form of agency will not provide the basis for a constitutivist vindication of those norms.

I suspect that a more-or-less explicit concern with heading off such a challenge underlies the widespread assumption that a satisfying constitutivism must accept Absolutism. Enoch offers an elaborate version of the challenge, but the basic idea is simple. Constitutivists seek to show that certain norms are unconditionally required by observing that anything you choose to do will be an action, and arguing that any action, as such, is subject to certain normative requirements. But, as Millgram puts it,

Hence it appears that a satisfying constitutivism must assume only the thinnest and most minimal reading of what an action is. But a pluralist constitutivist would be someone who accepts that there are several possible forms of agency, no one of which is necessarily exhibited by all subjects who can act for reasons. Hence a pluralist constitutivist would seem to deprive herself of the cornerstone on which constitutivist arguments are built: the premise that, necessarily, if you act at all, you must do something that is subject to such-and-such normative requirements.

I have described this challenge because I think it provides an intelligible motivation for the assumption that an acceptable constitutivism must be Absolutist. But while this motivation strikes me as intelligible, I think it actually rests on an unsound inference.

It is true that, if there is a plurality of forms of agency, then no one form of agency is necessarily exhibited by all subjects who can act for reasons. Properly understood, however, the constitutivist project need not rely on the claim that there is a form of agency that is necessarily exhibited by any subject who acts for reasons. The constitutivists’ claim need not be:

It might rather be:

On the latter reading, the starting-point of the constitutivist argument is not the claim that a certain form of agency is inescapable full stop, but that for a given subject, a certain form of agency is inescapable. If this is true, then that agent is inescapable subject to whatever normative requirements are implied by that form of agency, whether or not it is possible for there to be other kinds of agents subject to other kinds of norms. And that, in effect, is the pluralist's claim: not that there are norms to which any agent, simply qua agent, is subject, but that we, qua agents of the kind we are, are subject to certain specific norms.

Such a pluralist position will be defensible to the extent that it can be shown that there is a form of agency that is inescapable for us: not an optional refinement that we might choose to exhibit or not to exhibit, but a kind of self-determination that we, being the kind of beings we are, cannot fail to exhibit in our actions. If this condition were satisfied, then we could not coherently ask Enoch's question concerning this form of agency, even if other forms of agency were logically possible.

6. Reason and will: an Aristotelian approach

The preceding section concluded with an abstract description of a possible pluralist constitutivism. But could there actually be a position that met this description? Could we acknowledge that certain fundamental practical norms apply, not to every conceivable agent, but only to agents of a certain form, and yet maintain that, for agents of this form, these norms have an inescapable authority? I think Aristotle's ethics shows the possibility of such a position, and clarifies some of its crucial structural features. My aim in this final section is to sketch – very briefly and programmatically – a reading of Aristotle's ethics along these lines.

It will help to begin by recalling four very familiar features of Aristotelian ethics. First, Aristotle's approach is founded on an appeal to the idea that human beings, as human beings, have a proper work or a function.Footnote ²⁰ This, he holds, is what gives determinacy to the inquiry into what human happiness consists in, and thus into how we ought to live. The idea of a human function sounds alien to modern ears, but it becomes clear as the argument proceeds that what Aristotle has in mind is that human beings possess a specific kind of soul, one characterized by certain vital powers or principles, and chiefly, the power of reason. The function of man is to live (and thus to exercise the full complement of these powers) in a manner governed by reason, and Aristotle infers that, since the good of a thing is determined by its function, the human good consists in the excellent performance of those activities that belong to a reason-governed life.

This first point dictates the structure of Aristotelian ethics, but it does not yet determine its content. Although the main divisions of Aristotle's investigation are settled by his conception of the structure of the human soul, what falls within these divisions consists – and this is the second point – of an enumeration of various specific virtues. Aristotle makes no attempt to derive these virtues from his abstract characterization of the human function, and this has seemed to some commentators to make the initial appeal to function superfluous. This reaction, however, depends on a questionable view of the role that the notion of function is supposed to play. Apparently it is not supposed to provide a basis from which to derive particular virtues; Aristotle seems to regard it, rather, as a crucial clarification of what sort of thing a virtue is supposed to be. We might say that it is not the principle of their discovery, but the account of their metaphysical basis – of the kind of fact which can make it the case that a certain putative virtue is a genuine and not a merely pretended one.

Third, Aristotle presents his treatise on ethics as addressed, not to people needing to be convinced that the virtues he enumerates are virtues at all, but rather to those who have been ‘properly brought up’, brought up in such a way that they already are attracted to these ways of living.Footnote ²¹ Aristotle thus does not seek to address the sort of moral skeptic that Plato confronts in the person of Thrasymachus or Callicles: he merely develops a systematic framework in which to clarify and understand the unity of various goods to which his audience is already unreflectively drawn. And this modesty of justificatory ambition seems closely connected to certain other well-known features of Aristotle's view: his insistence that we should not expect the basic principles of ethics to admit of precise codification;Footnote ²² his likening of the understanding of the virtuous person to a capacity for accurate perception rather than for sound reasoning;Footnote ²³ and his identification of the case-by-case judgment of the practically wise person, rather than some abstract formula, as the standard of right action.Footnote ²⁴ All these features of Aristotle's view confirm the observation that he does not expect his general account of the human good to supply the principle of a non-question begging deduction or derivation of the soundness of certain ways of life, or of the right thing to do in particular circumstances. This is not to suggest that he would regard such questions as not open to argument, but he does not seem to aspire to offer a certain sort of foundational argument about such matters.

Finally, the fourth point: although the Nicomachean Ethics seeks to give a general account of human virtue, Aristotle is notoriously willing to contemplate the idea that different virtues are appropriate to a man, a woman, a child, and a slave. Whatever else we may make of this, it indicates at least that Aristotle thinks of his official system of virtues as characterizing, not norms applying to all possible agents possessing a rational soul, but norms appropriate to a specific sort of rational agent (namely, the patriarchs of free families living together in a political community). Moreover, in a striking passage, Aristotle remarks that, to the extent that we may think of other species of animals as having practical wisdom, it will be right to say that there is ‘a different philosophic wisdom about the good of each species.’Footnote ²⁵ The apparent implication is that the practical wisdom he characterizes in the Ethics is, not the only possible form of practical wisdom, but the kind pertaining specifically to human beings. We could express the significance of this by saying that, were there another species of rational animal, the fundamental principles of its practical thought might be quite different from those that govern ours. Perhaps the mode of excellent activity that characterized its reason-governed life would not require courage (think of a rational rabbit); perhaps it might not even be a specifically social rational animal, needing the virtue of justice. At any rate, our practical wisdom is the sound capacity for choice of a certain specific form of agent.Footnote ²⁶

I hope these descriptions of familiar aspects of Aristotle's position resonate with the case made earlier for the possibility of a pluralist constitutivism. Aristotle's ethics is evidently open to the possibility of pluralism, and he certainly does not aim to deduce his system of fundamental norms of action from an analysis of rational agency as such. He does not aim to deduce these norms at all, but simply (i) to give a clarifying description of them to people who already possess an incipient appreciation of their importance, and (ii) to articulate the metaphysical conditions that make it possible for this description to be sound or unsound.

In what sense is this a constitutivist position? I noted earlier that Aristotle's view can be regarded as constitutivist inasmuch as it seeks to ground the fact that we ought to act in certain ways in what we essentially are – where ‘what we essentially are’ includes, not just that we are rational agents, but that we are specifically human agents. It may seem, however, that this is a constitutivism in name only; for if it does not seek to show that the relevant sort of constitution has inescapable practical authority over those who bear it, and does not seek to deduce specific practical norms from this constitution, in what sense can it claim to ground these norms in our constitution?

At this point, it is important to recall what made the constitutivist approach attractive in the first place. It was, I take it, not its deductive or analytical aspirations per se, but rather what the relevant deductions and analyses were supposed to secure: an account of the source of fundamental practical norms that shows how they can be objective but not alien to our will, and that elucidates why their claim to authority over us is not one we can just shrug off, but one to which we are always already committed in the very exercise of our power to choose and act. The basic attraction of constitutivism, then, lies not in its being specially well placed to prove that certain fundamental practical norms apply to us, but in its giving an especially attractive account of what makes it the case that certain fundamental practical norms apply to us. Or to put it another way: the constitutivist's real contribution to metaethics consists, not in offering an account of the epistemology of normativity that would make its claims especially indubitable, but in offering an account of the metaphysics of normativity that would make the nature and authority of these norms particularly transparent.

It seems to me that Aristotle's approach to ethics can lay claim to these advantages of constitutivism in spite of its pluralist, non-analytical, non-deductivist character. What it offers is not a picture of how we can reason to the relevant practical norms; but that is not the basic sense in which the grounding of these norms needs clarification. What needs clarification is what sort of thing a ‘fundamental practical norm’ might be such that it could have the relevant combination of objectivity, authority, and capacity to move the will. Aristotle sketches a metaphysics that clarifies this by explaining the relation of the relevant norms to what we essentially are. His account aims to explain why, being the kind of agents we are, and receiving the kind of upbringing that such agents receive in favorable circumstances, we come to find the relevant norms primitively compelling – and also why (i.e., what makes it the case that) we are right to do so. A practically wise Aristotelian agent does not reason from his being (or being committed to being) a human being to specific norms of how he ought to live. His being a human being need not itself come into his practical thought at all, except incidentally. He faces no question about why he should be a human being, or do what human beings, as such, ought to do. It is enough that he is – essentially – a human being, and that this explains why it is sound for the goodness of certain specific ways of living and acting to figure primitively in his thinking about what to do and how to live.

I cannot develop the point here, but I suspect this aspect of Aristotle's ethics reflects much more basic features of his metaphysics. His metaphysics, famously, is hylomorphism: the view that the being of the most basic natural beings (natural substances) consists of a certain type of form being realized in a certain sort of matter. The essence of a natural substance consists, not of its mere form, but of such-and-such-a-form-in-such-and-such matter. In application to human beings, I would argue, this means that our essence is not simply: rational being, or rational animal, but rather: rational-animal-subject-to-such-and-such (specifically human) conditions. And our essence is that sort of being in the absence of we would not be at all. Hence it is not a mode of being one of us might fail to have. The consequence of this point, when brought together with Aristotle's project of explaining good in terms of essence, is that the measure of goodness that flows from our essence is not one that one of us can coherently reject, being what we are. This is not to say that the relevant measure of goodness is unquestionable, but that there is a true and intelligibly well-founded answer to such questions, an answer grounded in what can be called our constitution. We need not accept the substance of Aristotle's ethics to see the structural appeal of such a position. Why should a philosophically significant form of constitutivism claim anything more than this?