2. The Antiskeptical and Function Arguments

2.a. There are two main arguments for constitutivism. The first, antiskeptical, argument is that only constitutive standards “meet skeptical challenges to their authority with ease” (Korsgaard, Reference Korsgaard2009, 29; see also Ferrero, Reference Ferrero2009, Reference Ferrero2019; Katsafanas, Reference Katsafanas2013; Velleman, Reference Velleman2009, chap. 5). The second, function, argument is that constitutivism demystifies practical norms by making them continuous with other standards of goodness. Constitutive standards are grounded in the nature of the things they govern, so appealing to them need not require positing an independent, non-natural realm of normative facts. For instance, the standards of good eyesight and digestion are grounded in the nature and function of eyesight and digestion (Fix, Reference Fix2021, Reference Fix2023, Reference Fixforthcoming; Thompson, Reference Thompson2008, 28–30). Perhaps, too, what makes an action or agent good is determined by the nature of action or agency. Coming to know practical norms thus need not be any more mysterious than coming to know the standards of well-functioning for other capacities. They do not give rise to any special epistemological worry.

The two arguments depend on different senses of “constitutive standards.” The antiskeptical argument assumes that constitutive standards of action are transcendental norms. The function argument assumes that constitutive standards are functional norms. Sections 2.b–2.c develop these points in more detail. Section 2.d discusses the relationship between the two kinds of constitutivism: while they are in principle compatible, neither is a species or proper subset of the other.

Before continuing, let me clarify the function argument. The analogies to eyesight and digestion can make it seem like the argument supports Aristotelian naturalism of the sort defended by Philippa Foot (Reference Foot2001), according to which practical norms are grounded in our nature as specifically human beings. Plausibly, the standards of well-functioning for eyesight and digestion depend on their role in their bearer’s life-form or species. So perhaps standards of well-functioning for action are also life-form-specific.Footnote ⁵ Is the distinction between transcendental and functional constitutivism then just a distinction between transcendental constitutivism and Aristotelian naturalism? No. Functional constitutivists can but need not be Aristotelian naturalists. The function argument is meant to establish that the standards of goodness for an activity are grounded in its nature. That the standards of goodness for intentional action itself are life-form specific is a further, substantive premise—one functional constitutivists can deny. Smith (Reference Smith2015, 193–195), for instance, thinks that we have a function as agents, and that this function can come apart from our function as human beings.Footnote ⁶ Further, one can be a transcendental constitutivist who thinks practical norms are grounded in specifically human psychology: all human agents are committed to complying with some norm on pain of inconsistency.Footnote ⁷ The distinction between transcendental and functional constitutivism thus cuts across the distinction between grounding practical norms in our rational nature and grounding them in human nature.

2.b. Start with the antiskeptical argument. As quoted above, Korsgaard (Reference Korsgaard2009, 29) says that constitutive standards “meet skeptical challenges to their authority with ease.” Elsewhere she and others put the same point in different terms: “if we can identify something as an internal norm [that is, a constitutive standard], the question why you should conform to the norm answers itself” (Korsgaard, Reference Korsgaard2008b, 61); “questioning” a constitutive aim “simply doesn’t make sense” (Katsafanas, Reference Katsafanas2013, 59); “the norm of correctness for action should be exempt from deliberative criticism … the norm should not leave open any question about whether to act in accordance with it” (Velleman, Reference Velleman2015a, 28). The general idea is that if a norm is a constitutive norm of rational agency, then it silences skeptical doubts. Further, only such norms silence skeptical doubts, since any other norm could be rejected without inconsistency (Korsgaard, Reference Korsgaard2009, 32–33). And a norm is genuinely authoritative—I ought to comply with it—if and only if it silences skeptical doubts. Thus, only constitutive norms of rational agency are genuinely authoritative.Footnote ⁸

The first two premises are true just in case constitutive norms are transcendental norms. If I’m rationally committed to a norm on pain of inconsistency, then it silences skeptical doubts. Any question of the form “Why should I comply with N?” can be answered just by showing that I must comply with it by my own lights. And if I’m not committed to a norm on pain of inconsistency, then it does not silence skeptical doubts; “Why should I comply with N?” questions cannot be given answers I must accept on pain of inconsistency. So if constitutive norms aren’t transcendental norms, then I’m not committed to complying with them on pain of inconsistency. Then, they do not survive skeptical questioning: I might coherently refuse to comply with them.

2.c. Constitutivists who think constitutive norms are functional but not transcendental norms thus cannot appeal to the antiskeptical argument.Footnote ⁹ But they can appeal to the function argument. It goes like this. Standards of goodness for a thing are typically determined by its function: what makes for a good house or healthy lion depends on the function of houses and lions. The point is more vivid when we consider exercises of animal capacities: what counts as good vision or digestion is determined by vision’s or digestion’s respective functions. Since the capacity to act is just another animal capacity, the same should go for action: what counts as a good action is determined by action’s function. A good action is one that realizes the function of action. And we ought to act well, in the functional sense of “well.” Thus, we ought to do what’s necessary to realize the function of action. What we ought to do depends on the function of action. In premise-conclusion form:

1. (1) A good instance of a kind K realizes the function of Ks.

2. (2) An action Φ-ing realizes the function of action only if it complies with norm N.

3. (3) Thus, Φ-ing is a good action only if it complies with norm N.

4. (4) One ought to Φ only if Φ-ing is a good action.

5. (5) Thus, one ought to Φ only if Φ-ing complies with N.Footnote ¹⁰

The function argument fits practical normativity into a broader picture of the nature of normativity. Practical norms are no more mysterious than the standards for good houses or vision. Like other standards of goodness, they are determined by the functions of the things they govern.

An upshot of the discussion so far is that transcendental constitutivism promises something (merely) functional constitutivism cannot, but at a steep theoretical cost. It promises to silence skeptical worries. Merely functional constitutivism has many virtues, but silencing skeptical doubts is not one of them. The theoretical cost is that to argue a given norm is a constitutive standard of rational agency, transcendental constitutivists must show that anyone who flouts that norm is not just mistaken but inconsistent. Transcendental constitutivists thus set a high—essentially Cartesian—argumentative bar for themselves. By contrast, functional constitutivism does not commit its defenders to any particular way of arguing that some norms are constitutive standards.

Here’s another stab at the same thought. Constitutivism as such is an explanatory claim: practical norms are grounded in the nature of action. Transcendental constitutivists say something further: we can only argue that some norm is authoritative by deriving it—using just the requirement of consistency—from the nature of action. In Kant’s (Reference Kant and Gregor2015) parlance, they make claims about both the ratio essendi and the ratio cognoscendi of practical norms. By contrast, functional constitutivists just make a claim about their ratio essendi. In this respect, transcendental constitutivism is more committal than functional constitutivism.

2.d. Is transcendental constitutivism, then, just a species of functional constitutivism? No. They are compatible but separable in both directions. They’re compatible because one might think that for something to be the function of action just is for any rational agent to be committed to it on pain of inconsistency.Footnote ¹¹ Then, functional norms are transcendental norms. But functional constitutivists need not be transcendental constitutivists because they needn’t say an agent is necessarily committed to realizing her function on pain of inconsistency. Perhaps all we can say of an agent who does not realize the function of acting is that she does not act well. And transcendental constitutivists need not be functional ones because they might similarly think that the function of action and what any agent is committed to come apart—and that only the latter matters.

To see the latter possibility, imagine a transcendental constitutivist who agrees with Smith’s or Foot’s theories of well-functioning action. She agrees that the function of action is to realize the agent’s desires or is determined by the role of action in the agent’s life-form, respectively. Still, she might ask: why care about that? What is it to her if she is ill-functioning in Smith’s or Foot’s senses? Indeed, this sort of question is prevalent in the literature on Aristotelian naturalism in the form of the “Why be a good human being?” objection (Lott, Reference Lott2014).Footnote ¹² A transcendental constitutivist pressing this question agrees that action has a function—just like vision and digestion do—but also thinks that realizing that function can come apart from acting as one is committed to on pain of inconsistency. They thus reject premise (4) of the above argument: they think that one ought to comply with the norms to which one is committed on pain of inconsistency, not to perform functionally good actions.

Indeed, we can make things more extreme: a transcendental constitutivist might deny that action has any function at all. Granted, any plausible form of transcendental constitutivism will say that transcendental norms are grounded in the nature of action or agency. But this does not imply that action has a function—especially if “well-functioning” is meant to be univocal across “the function of digestion,” “the function of vision,” and “the function of action.” There might be some norms with which we are committed to complying on pain of inconsistency, even if those norms are not relevantly analogous to the functional ones governing digestion, vision, and other human capacities. Transcendental constitutivists thus need not be functional ones.

3. The Shmagency Problem

3.a. The heart of the shmagency problem is that constitutivists cannot answer the question: “Why does the fact that N is a constitutive standard of agency mean that I should comply with N?” (Enoch, Reference Enoch2006, Reference Enoch and Brady2011, Reference Enoch2020). Given the distinction between concepts of constitutive standards, there are really two questions—one for each sort of standard. For transcendental constitutivists, the question is:

By contrast, for functional constitutivist, it’s:

At the very least, we should acknowledge that these are different questions. Given that the extant literature on constitutivism has largely treated the shmagency problem as one unified problem for constitutivists, that alone might be surprising. Further, I’ll argue that the most plausible answers to each question are different. In broad strokes, they both involve offering what I’ll call “identificatory” theories of practical norms, of the form:

Identificatory theories state metaphysical identities, like “to be water is to be H₂O” or “to be a fox is to be a female vixen.” The underlying thought is that for constitutivists, for something to be an authoritative practical norm is for it to be a constitutive standard of action.

I use the cumbersome term “identificatory” rather than the more familiar “reductive” for a few reasons. First, talk of “reduction” is often associated with a claim about explanatory priority: if A is reducible to B, then B is more fundamental than A.Footnote ¹³ On this reading, reduction is asymmetric. By contrast, identification is symmetric: if for p to be the case is for q to be the case, then for q to be the case is for p to be the case.Footnote ¹⁴ Second, some use “reductive”—or at least “reductive naturalism”—in metaethics to mean that the thing to which practical norms (or moral properties, etc.) are reduced “would be countenanced in a scientifically constrained view of what exists” (Copp, Reference Copp and Kirchin2017, 31). I take no stance on whether constitutive standards (in either sense) are countenanced in a scientifically constrained view of what exists.Footnote ¹⁵ Third, talk of reduction in metaethics can give the impression that the thing reduced does not really exist: according to reductive views, it turns out that there are no normative facts.Footnote ¹⁶ I think this impression is misleading across the board, but best to nip it in the bud. “Identificatory” does better on this front: an identificatory theory of practical norms does not say practical norms do not exist any more than the theory of water, according to which to be water is to be H₂O says that water does not exist.

3.b. Now for the details. Start with transcendental constitutivism. Suppose all rational agents are committed to complying with some norm N on pain of inconsistency. Enoch asks: what’s that got to do with having a reason to comply with N? The accusation is that constitutivists illicitly infer an “ought” from an “is.” The fact that we are committed to complying with some norm does not mean we ought to comply with it. Indeed, Enoch describes the constitutivist strategy as “an especially problematic instance of” the naturalistic fallacy—an illicit move “from someone caring about something to it being the case that she should care about it” (Enoch, Reference Enoch and Brady2011, 215).

To foreshadow my position: Enoch is right that if the naturalistic fallacy is a fallacy, then transcendental constitutivists (and functional ones—as we’ll soon see) commit it. They infer an “ought” from an “is.” But constitutivists of both sorts should deny that it’s a fallacy, let alone that the constitutivist inference is “especially problematic” instance of it. Indeed, they should say that it is a uniquely unproblematic instance of it.

To get a grip on the problem, consider Barry Stroud’s (Reference Stroud1968) discussion of transcendental arguments. Considering arguments that some proposition S is a necessary condition of the possibility of language, Stroud says that “the skeptic can always very plausibly insist that it is enough to make language possible if we believe that S is true, or if it looks for all the world as if it is, but that S needn’t actually be true” (Stroud, Reference Stroud1968, 255). The shmagency problem applies this reasoning to constitutivism: perhaps we are committed to complying with N, or we must believe that we ought to comply with N, but that does not imply that we actually ought to do so.Footnote ¹⁷ There might be a gap between what we are committed to and what’s true or right.

Korsgaard’s “transcendental argument” for the value of humanity in Sources makes the analogy apparent:

The last inference looks invalid. Perhaps we necessarily find ourselves to be valuable (or are committed to believing ourselves valuable, etc.). But why does that imply that we are valuable? Korsgaard seems to illicitly move from “is” to “ought,” and from a claim about what we must believe to a claim about what’s true.

Transcendental constitutivists need the missing premise: if we necessarily find ourselves valuable, then we are valuable. In my terms: if we are necessarily committed to complying with N, on pain of inconsistency, then we must comply with N. How do transcendental constitutivists defend that premise?

A popular strategy is to identify its being the case that one must (or ought to) do something with its being the case that one is committed to doing it on pain of inconsistency. Call this position “Constructivism”:

Constructivism entails the missing premise by identifying its antecedent with its consequent. Being rationally required to comply with a norm just is being committed to it on pain of inconsistency. Constructivists bridge the is-ought gap by denying it exists when the relevant “is” is being committed to a norm on pain of inconsistency. And Constructivism provides a straightforward answer to the question of why the fact that one is committed to doing something means that one ought to do it: those are the same thing. It’s like asking why the fact that something is water means that it is H₂O.

Korsgaard (Reference Korsgaard and O’Neill1996, 92–98, Reference Korsgaard2008a, 321–324) defends Constructivism by appealing to an account of normative concepts. Rational agents face the practical problem of how to act. Normative concepts like “ought,” “right,” and “must” are placeholders or names for the solution to this problem. To say that Φ-ing is right or that I must Φ is to say that Φ-ing is a solution to the practical problem of how to act. The task of moral philosophy, then, is to move from normative concepts like “must” and “right” to normative conceptions—moral theories telling us which actions are right. If I’m necessarily committed to Φ-ing on pain of inconsistency, then I have no consistent way of settling the question whether to Φ negatively. There is no consistent route to not-Φ-ing, so I’ve solved the problem of whether to Φ. So, I must Φ.

This is the sense in which practical normativity is “essentially first-personal” for constructivists. Ethical thought is, in the first place, thought about what to do—not about an independent realm of facts. If we are inescapably committed to complying with some norm on pain of inconsistency, then we can only consistently settle the question of whether to comply with that norm affirmatively. For transcendental, constructivist constitutivists, that’s all it takes for it to be the case that one ought to comply with it.

Still, something seems fishy about the appeal to Constructivism. Transcendental constitutivists set out to silence skeptical doubts. But it seems like a shmagent can raise skeptical doubts about Constructivism itself, so it just pushes the problem back. Why is Constructivism true? Why is being rationally required to comply with a norm the same as being committed to doing so on pain of inconsistency?

At this point, the sense of “why” at issue matters. If the question asks for a metaphysical explanation—“in virtue of what is Constructivism true?”—then Constructivists should reject the question. It is like asking “why is water H₂O?”. There is no answer, and that’s no problem, as metaphysical identities have no explanations.Footnote ¹⁹

But the question might instead be: “why believe Constructivism is true?”Footnote ²⁰ That is an intelligible question, just like “why believe water is H₂O?” Perhaps Korsgaard’s argument for Constructivism, or some other one, answers it. But that argument does not show that disbelieving Constructivism is inconsistent; just that it’s erroneous. A skeptical interlocutor might be unconvinced of it without embroiling herself in any obvious inconsistency (indeed, many have been!).Footnote ²¹ So it looks like transcendental constitutivists still have a problem: they cannot live up to their own antiskeptical ambitions. Even if there are norms to which we are committed on pain of inconsistency, a skeptic can intelligibly ask whether she should comply with them. That is the heart of the shmagency problem, and it’s why transcendental constitutivists should take it seriously.

Whether the problem is decisive—which I will not settle here—depends on which skeptical doubts must be silenced. If all skeptical doubts of all forms must be silenced, then transcendental constitutivism is hopeless. Transcendental constitutivists will be left with no ground from which to argue. For even if we were committed to believing Constructivism on pain of inconsistency, a skeptic might still reject the requirement to be consistent.Footnote ²² But if only some skeptical doubts must be silenced, transcendental constitutivists owe us an account of which. Which conclusions can we only defend by showing that one is inconsistent if one does not accept them, and which conclusions can we defend by other means?Footnote ²³

3.c. The functional constitutivist response to the shmagency problem is more straightforward. Functional constitutivists infer “rational agents must comply with N” from “the function of action is to comply with N.” Proponents of the shmagency worry ask: “why does the fact that the function of action is to comply with N imply that rational agents must comply with N?”

Like transcendental constitutivists, functional constitutivists should accept an identificatory theory of practical normativity. Call it “Functionalism”:

Like Constructivism, Functionalism solves the shmagency problem by denying that there is a metaphysical is-ought gap, this time where the relevant “is” is being the function of rational action. Appealing to Functionalism might seem like cheating. The point of the shmagency problem, one might say, is that constitutivists want an answer to the question “why must I comply with this norm?” Faced with the answer “doing so is the function of action,” a skeptical shmagent can re-raise the skeptical worry: “But I do not care about realizing the function of action. Why must I?”

Two answers—one clarificatory and one substantive. The clarificatory one: functional constitutivists do not say that we should care about the function of action, under that description. They rather say that we must comply with N. The reason why we must do so is that doing so is necessary to realize the function of action. But “complying with N is necessary to realize the function of action” is not itself a normative reason to comply with N, like “Bob is in pain” is a normative reason to help Bob.Footnote ²⁵ Functional constitutivists should thus deny the presupposition of the question “Why must I care about realizing the function of action?” The relevant question in the vicinity is: “Why must I comply with N? I get that doing so is the function of action, but I can consistently refuse to realize the function of action. So what more do you have to say to me?” This is Enoch’s (Reference Enoch2006, 42) worry that threatening wrongdoers with failure to realize the function of action is not a “threat that will strike terror into the hearts of the wicked.”Footnote ²⁶ It’s also a skeptical worry: what more can be said to convince someone skeptical of complying with N to do so? In fact, it’s the exact skeptical worry that I earlier imagined an antiskeptical constitutivist who accepts Smith’s or Foot’s accounts of the function of action raising.

Now the substantive answer: functional constitutivists should concede that they cannot answer all skeptical doubts and deny that it matters. The appeal of functional constitutivism is not that it silences skeptical doubts, but that it naturalizes practical norms by making them functional ones—promising ethical objectivity without the problems of non-naturalist realism. Indeed, the considerations that support Functionalism are more-or-less the same ones that motivated the function argument. Insofar as the shmagency problem is just that functional norms do not meet skeptical challenges, there’s nothing to worry about. There’s a huge gulf between functional and transcendental constitutivists here. Transcendental constitutivists go in for constitutivism precisely because constitutive standards silence skeptical doubts. Conceding that constitutivists cannot answer skeptical questions would, for them, be defeat. For functional constitutivists, it’s no problem.

We could instead understand the shmagency problem as a version of Moore’s open question argument (OQA).Footnote ²⁷ Understood this way, the shmagency argument is an objection to Functionalism. “Is being necessary to realize the function of action the same as being required?” is an open question. The answer is not contained in the question (unlike the answer to “is being a bachelor the same as being an unmarried man?”). But if being required just is being necessary to realize the function of action (i.e., if Functionalism is true), then the question should be closed. Anyone who understands the question properly will have to recognize the identity. So Functionalism is false.

Smith (Reference Smith2015, 198) says that if the shmagency argument is an OQA, then “we should simply take note of [Enoch’s] Moorean intuitions and move on.” That response is a bit hasty, but its spirit is right. In general, OQAs help distinguish concepts, not properties. The fact that “Is water H₂O?” is an open question shows that water and H₂O are different concepts, not that being water and being H₂O are different properties. But Functionalism is a theory of normative properties, not concepts. The function argument is about what it is for an action to be good and for it to be the case that one must perform an action, not about the concepts good and must. And constitutivism avoids metaphysical and epistemological mystery insofar as practical norms just are functional ones, not insofar as normative concepts are functional ones. OQAs thus do not pose a problem for Functionalism.Footnote ²⁸

Functional constitutivists thus should not worry about the shmagency problem. The problem is either an explanatory or normative demand (the intelligibility of which Functionalists deny), a skeptical question (which functional constitutivists were not trying to answer in the first place), or an open question argument (which functional constitutivists should deny is sound).

3.d. Both transcendental and functional constitutivists should offer identificatory theories of practical normativity. Enoch (Reference Enoch2020, 342) presses a last line of response: offering an identificatory theory robs constitutivism of its special appeal. “[C]onstitutivists,” he says, “never thought of themselves as plain-old naturalists.” Rather, they aim to offer “a really special kind of naturalist reduction”—one that fares better than other naturalist theories (Enoch, Reference Enoch2020, 342, emphasis his). Enoch’s response illustrates the distinction between transcendental and functional constitutivism, so I’ll quote it in full:

Three threads are worth pulling apart. First, does constitutivism avoid “objectionable externalism or alienation” (by securing the right kind of connection to the agent’s desires)? Second, does it “close questions that on other reductions remain open”? Third, why think that the goodness-fixing kind agent is related to reasons while toaster is not? The answer to each question depends on whether we are talking about functional or transcendental constitutivism.

First thread first. Some constitutivists appeal to something like Williams-style internalism to motivate constitutivism (Katsafanas, Reference Katsafanas and Star2018, 379–381; Korsgaard, Reference Korsgaard2008b; Velleman, Reference Velleman2009, chap. 5, Reference Velleman2015b; Williams, Reference Williams1981). Here’s a rough definition of internalism:

Internalism lends support to transcendental constitutivism. If I should comply with a norm only if I can deliberate from my current motivational set to complying with the norm, then (perhaps only) transcendental constitutivism can secure ethical objectivity. For if I’m necessarily committed to complying with N on pain of inconsistency, then I can reason to doing so from my subjective motivational set no matter its content. If I’m not, then I might have a motivational set from which I cannot reason to a decision to do something which complies with N, and so might not be subject to N. So if externalism is indeed objectionable, transcendental constitutivists avoid it.

The connection between functional constitutivism and internalism is trickier. They are compatible but separable. Perhaps complying with N is the function of agency, but my current motivational set is so defective that I cannot comply with it.Footnote ³⁰ Still, functional constitutivism might avoid alienation. After all, it’s clear what practical norms have to do with me on the functionalist account: they are grounded in my nature (Lavin, Reference Lavin2017; Schafer, Reference Schafer2023, 115). They’re not imposed on me from without, and in this sense, I’m not alienated from them. Even an identificatory form of functional constitutivism can thus avoid “objectionable externalism or alienation.”

Second, do constitutivists aim to close open questions? Transcendental constitutivists do—at least in that they aim to silence skeptical doubts. Functional constitutivists, by contrast, do not—as I argued in §3.c.

Third, are constitutivists “better positioned than other naturalists [or identificatory theorists] vis-à-vis Moorean worries”? Transcendental constitutivists do seem better positioned, precisely because they can appeal to internalism and claim to silence skeptical doubts. But functional constitutivists neither appeal to internalism nor aim to silence skeptical doubts. Are they in the same boat as every other identificatory or “reductive” naturalist, then? No! They can appeal to the function argument; most other identificatory theorists cannot. One might still ask: why think there’s a special connection between practical norms and the function of intentional action—rather than the function of something else, like toasters? With toasters, the answer is easy: because practical norms govern rational agents and their actions, not toasters and toasting.Footnote ³¹ A more pressing worry is: why think there’s a connection between practical norms and the function of rational action rather than, say, the function of a human being or something else? The most straightforward answer appeals to the “compelling thought” that “judgments of practical reason are assessments of agency, not some other aspects of our lives”: to say someone acted well or poorly is to evaluate her as rational agent, not as anything else (Setiya, Reference Setiya2017b, 240).Footnote ³² Indeed, this is supported by the initial presentation of the function argument: what counts as digesting well depends on digestion’s function, what counts as seeing well depends on what vision’s function, and so on. So too, what counts as acting well should depend on action’s function—not on the nature of something else.

3.e. My aim is not to defend either version of constitutivism wholesale against the shmagency problem here—though hopefully I’ve shown how the argument(s) should go. My aim is to show that which kind of constitutivism we are considering makes a huge difference to how constitutivists should respond, and to how seriously they should take the problem. Transcendental and functional constitutivism are distinct—and the distinction makes a difference.

4. The Bad Action Problem

4.a. The bad action problem says that constitutivism makes bad action impossible.Footnote ³³ To get an intuitive sense of the problem, consider that constitutivists sometimes say things like:

1. (a) The instrumental principle is constitutive of an act of the will. If you do not follow it, you are not willing the end at all. (Korsgaard, Reference Korsgaard2008b, 56)

And:

1. (b) If particularistic willing is impossible, then when you will a maxim you must take it to be universal. If you do not, then you are not determining yourself to perform an action, and then you are not willing. (Korsgaard, Reference Korsgaard2009, 76)

If practical norms are constitutive norms of willing—as Korsgaard thinks—then these passages make it seem like the instrumental principle and the requirement not to will particularistically are inviolable. If you do not comply with them, then you are not willing at all.

But then how can they set normative standards? A practical norm is authoritative only if one can fail to comply with it.Footnote ³⁴ If it’s impossible for me to fail to take the means to my ends, then the rule “take the means to your ends” is not authoritative. Since anything I do will comply with it, it cannot guide my conduct.

We can state the problem in premise-conclusion form:

1. (1) If constitutivism is true, then rational agents must comply with the constitutive norms of action.

2. (2) Necessarily, all actions comply with the constitutive norms of action.

3. (3) An agent must comply with a norm N only if it’s possible for her to violate N.

4. (4) Thus, constitutivism is not true.

(1) follows from the definition of constitutivism. (3) says practical norms must be violable. Constitutivists must take issue with premise (2).

How we understand premise (2) depends on the sense of “constitutive norm” at issue. For transcendental constitutivists, it’s:

By contrast, for functional constitutivists, it’s:

The rest of this section discusses (2T) and (2F). I argue that (2T) is false, so transcendental constitutivists needn’t worry about the bad action problem as stated. But still, I will argue, they face a variation of the bad action problem. For functional constitutivists, it’s more complicated. (2F) requires further specification: its plausibility depends on what it is for something to be the function of action. The best version of functional constitutivism, I argue, renders (2F) false and thereby avoids the bad action problem.

4.b. (2T) is false: it’s possible for an action to violate a norm to which the agent is committed on pain of inconsistency. Indeed, there is nothing mysterious about failing to comply with a norm to which one is committed on pain of inconsistency. We’re inconsistent all the time! So the bad action problem is no problem at all for transcendental constitutivists.

Consider a doxastic case. If I believe p and if p, q, then I’m rationally committed to the truth of q. Still, I might fail to believe it. Indeed, this happens whenever I do not believe all the logical consequences of my beliefs. One might worry that believing p and if p, q commits me to the truth of q, but not to believing q. So I do not fail to uphold any doxastic commitment in not believing q. But we can modify the example. Believing p and if p, q commits me to not believing not-q. Believing p, believing if p, q, and believing not-q at the same time is inconsistent. But it’s nevertheless possible—we have inconsistent beliefs without realizing it all the time. We can just fail to put the pieces together. Bad action, for transcendental constitutivists, is the practical analogue of failing to coherently put one’s beliefs together.Footnote ³⁵ There is no reason to expect it to be impossible.

This is, at least sometimes, Korsgaard’s solution to the bad action problem. She slides between saying things like (a) and (b) from earlier, and saying things like:

1. (c) [W]illing an end is equivalent to committing yourself, first personally, to taking the means to that end. (Korsgaard, Reference Korsgaard2008b, 57, emphasis mine)

And:

1. (d) Unless we are guided by these principles [the hypothetical and categorical imperatives]—unless we are at least trying to conform to them—we are not willing or acting at all. (Korsgaard, Reference Korsgaard2009, 81, emphasis mine)

While (a) and (b) make the bad action problem look pressing, (c) and (d) do not. Failure to comply with the instrumental principle or categorical imperative is no more mysterious than failure to uphold any other commitment.

Still, transcendental constitutivists face a related worry. I’ll call it the “escapability problem.” If moral norms are transcendental norms, then everything subject to moral standards must be subject to such norms. Transcendental norms are such that in violating them, I act inconsistently. But transcendental constitutivists typically derive moral norms not from just any action but from action performed for a non-instrumental (“final”) reason (e.g., Korsgaard, Reference Korsgaard and O’Neill1996, chap. 3). In acting immorally for a final reason, I act inconsistently. Suppose, then, I can act for a merely instrumental reason—in order to do something else but for no final reason. For instance, I can reach for a snack to eat it but do not aim to eat it for any reason—I’m just determined to do so by my desires. Transcendental constitutivists must say reaching for the snack—and anything else I do just for an instrumental reason—is not subject to moral standards; it cannot be immoral. But surely if we can exercise merely instrumental agency, we can do so immorally! If desire determines me to eat the snack and I steal yours to do so, I act wrongly.

The escapability problem is a version of the bad action problem. The bad action problem says that constitutivism is incompatible with any action’s being bad. The escapability problem says that constitutivism is incompatible with some action that is indeed bad—like stealing your snack merely from desire—being bad. The escapability problem also resembles Enoch’s (Reference Enoch2006) original presentation of the shmagency problem. His shmagent asks: why can’t I be a shmagent, who’s just like an agent but does not share the constitutive aim of agency? The character I’m considering asks: why can’t I act for just instrumental reasons and thereby exempt myself from the transcendental norms of non-instrumental agency?

Korsgaard solves the escapability problem by denying that we can act just for instrumental reasons—that we can do one thing to do another without acting for a final reason. Consider the oft-cited passage from Sources: a rational agent “must endorse the desire before it can act on it, it must say to itself that the desire is a reason” (Korsgaard, Reference Korsgaard and O’Neill1996, 94). Her aim is to rule out an agent who is moved by desire without acting for a final reason. She endorses:

Finality solves the escapability problem by saying that acting for a final reason is, in the relevant sense, inescapable.Footnote ³⁷ All actions done for a reason are done for a final reason and so are subject to the constitutive standards of acting-for-a-final-reason. And any action not done for a final reason is not done for a reason at all and so is outside the scope of transcendental constitutivists’ aims. But Finality is controversial and not obviously plausible: intuitively, it seems like I can be moved by desire to pursue an end for no final reason.Footnote ³⁸ So while transcendental constitutivists avoid the standard bad action problem, they replace it with the related, perhaps equally worrying escapability problem.

4.c. For functional constitutivists, the bad action problem is less well-defined. Recall the key premise:

The truth of (2F) depends on what it is for something to be the function of action. One possibility is that realizing the function of action is what makes a behavior into an action. That is, if something is an action then it is so because it realizes the function of acting. This implies that anything that fails to realize the function of action is not an action after all, so (2F) is true.

Do constitutivists really say this? Some do. In describing his “functionalist” view, Velleman (Reference Velleman2022, 86) “identifies normativity with whatever…transforms what would otherwise be impulsive, unwitting, or neurotic behavior into human action.” Elsewhere, he says that the constitutive standards of action are set by the nature of “[f]ull-blooded human action” (Velleman, Reference Velleman2015a, 15). The properties of full-blooded action are those that all human actions ought to have. Korsgaard is even more explicit:

For her, conformity to the categorical imperative is what makes something into an exercise of my will—a genuine act of self-determination—rather than a mere episode of my desire causing me to move. Anything violating the categorical imperative, then, is not an action after all but rather a mere movement determined by desire.Footnote ³⁹

The bad action problem seems pressing, I think, because constitutivists like Korsgaard and Velleman say things like this, and so they must do some fancy footwork to avoid the problem. They say action itself comes in degrees: something is an action to the extent that it realizes the function of acting (Korsgaard, Reference Korsgaard2009, chap. 8; Velleman, Reference Velleman2015a). A defective action is not just a bad action; it’s less of an action. But this does not seem plausible. Put plainly, it sure seems like someone can do something that is an action in every sense and yet still wrong.Footnote ⁴⁰

Korsgaard and Velleman’s mistake is saying that if something is an action, then it is so because and to the extent that it realizes the function of acting. In general, what makes something a member of a functional kind is not that it succeeds in realizing its function: members of functional kinds can function poorly without ceasing to be members altogether. Nor is something a member of a functional kind just to the extent that it realizes its function. Ill-functioning toasters and bad digestion are possible, and they are toasters and digestion no less. So it’s not the case that successfully toasting or digesting is what makes something a toaster or an instance of digestion. Perhaps something can so egregiously fail to toast or so completely fail to digest that it is not a toaster or instance of digestion after all. But even that does not imply that defective instances are somehow less instances (e.g., that defective digestion is digestion to a lesser extent than successful digestion). Why should action be any different?

Functional constitutivists should say that action is just like any other functional kind. What makes something an action differs from what makes it a good action. Individual actions are good to the extent that they realize the function of action, just as individual toasters are good to the extent that they realize the function of toasters, and instances of digestion are good to the extent that they realize the function of digestion. Admittedly, there remains a challenge here: functional constitutivists must give a positive account of the sense in which something can be “constitutive” feature of action without being a necessary property of all actions. But there is a principled reason to think this challenge can be met: it’s met in the case of other functional kinds. So violable standards of goodness for something can, in principle, be grounded in that thing’s nature.Footnote ⁴¹ Functional constitutivists must take care to avoid the bad action problem, but they can avoid it.

What about the escapability problem? Must functional constitutivists endorse Finality? It depends on whether one can fail to realize the function of acting-for-a-final-reason without acting for a final reason. And that will depend on what the relevant function is. If the constitutive function of acting for a final reason is to unify the agent (as in Korsgaard, Reference Korsgaard2009), then I can fail to realize it without acting finally for a reason.Footnote ⁴² There would be two ways to fail: I could fail in finally acting for a reason or fail to finally act for a reason. In terms of capacities and their exercises: both failures in exercising the capacity or fail to exercise it manifest defects of the capacity.Footnote ⁴³ But consider another activity: walking. I walk badly if I “trip, slip, stumble, or tumble” (Fix, Reference Fix2020, 41). But I can walk badly only if I’m walking—or at least trying to walk. I do not fail to realize the function of walking when I spend my evening rewatching Barbie rather than going for a stroll. Failing to walk does not mean failing to realize the function of walking. If acting for a final reason is like walking, then functional constitutivists must also endorse Finality. But if one can fail to realize the function of acting for a final reason when one is not even trying to do so, then they needn’t. Since I have not said anything about the function of acting-for-a-final-reason, I cannot settle this matter here. The most I can say is that transcendental constitutivists need Finality, while functional constitutivists might be able to do without it. In this respect, functional constitutivists are better positioned to solve the escapability problem.

4.d. I’ve not solved the bad action problem. But I’ve disambiguated two worries. Transcendental constitutivists do not face the bad action problem as it’s often presented, but they must defend Finality to solve the escapability problem. By contrast, functional constitutivists may not need to defend Finality, but they must tread carefully to avoid the traditional bad action problem.

5. Conclusion: Two Debates Rather Than One

Constitutivism is a genus of metaethical views. It is unified by the thesis that practical norms are grounded in our nature—that they are constitutive standards of action or agency. Its species are distinct because they disagree about how norms are grounded in our nature—about what it is for a norm to be a constitutive standard of action or agency.

Some objections to constitutivism are, pitched at a suitably general level, objections to all of its species. For instance, both functional and transcendental constitutivists should endorse identificatory theories of practical normativity. Arguments that such theories are doomed to fail are objections to constitutivism as a genus. Similarly, arguments that the nature of action is “too thin” to ground any practical norms—transcendental or functional—spell problems for constitutivists of all stripes.

But really, the debate is in the details. The arguments for constitutivism vary by species. Transcendental constitutivists can appeal to the antiskeptical argument; functional constitutivists can appeal to the function argument. And the virtues of constitutivists’ identificatory accounts of normativity depend on which account is at issue. So while the debate over constitutivism might start at a high level of generality, it cannot end there. In the end, transcendental and functional constitutivism have different virtues and face different objections. There are two debates here, not one. They should proceed in parallel, not in tandem.