2. Kantian Constructivism

Kantian constructivism is by now a well-established position in metaethics. The term was introduced by John Rawls in his Reference Rawls1980 paper ‘Kantian Constructivism in Moral Theory’. In contrast to metaethical realism, constructivism claims that there are no moral facts which exist objectively, in the sense of independently of human beings. Moral norms are constructed by human beings, under certain specified conditions, and it is these conditions which give them their normative validity. As Rawls puts it, ‘Kantian constructivism holds that moral objectivity is to be understood in terms of a suitably constructed social point of view that all can accept. Apart from the procedure of constructing the principles of justice, there are no moral facts’ (Rawls Reference Rawls1980: 519). Rawls argues that ethics is not, as realists assume, a branch of epistemology, and its procedures are not discovery procedures, aiming to bring to light pre-existing moral facts. Ethics is practical, and its procedures construct moral norms which would not exist, or would not have validity, otherwise (Rawls Reference Rawls1980: 523–4).

What makes constructivism Kantian, for Rawls, is that it relies on certain features of human beings emphasized in Kant’s moral philosophy: our status as rational autonomous agents, self-legislators who in the light of reasons impose moral norms on ourselves, these norms being valid because we have imposed them on ourselves, having first constructed or tested them by following valid procedures (Rawls Reference Rawls1980: 520–1). Rawls’ construction procedure, the original position, constrained by the veil of ignorance, is designed to isolate and draw out these features of human beings (Rawls Reference Rawls1980: 549–50).

Kantian constructivism has been developed further by Christine Korsgaard (Reference Korsgaard1996, 2008, 2009), Onora O’Neill (Reference O’Neill1989), Carla Bagnoli (Reference Bagnoli and Carla2013, 2015) and others. Constructivism claims to be ‘a third, distinct possibility’, ‘somewhere in the space between realist and relativist accounts of ethics’ (O’Neill Reference O’Neill1989: 206). It carves out a position in between metaethical realists, who claim that there are objective moral facts, and relativists (broadly construed), who see morality as entirely subjective.

Constructivism varies in its scope: it may be global or local. A global constructivist would hold that all normativity is constructed. This is a very strong position, runs into evident problems, and has found few adherents (Enoch 2009; Wallace 2012: 21–2). Although critics sometimes class Kantian constructivism as global, this is not borne out by the kind of norms which the latter concentrates on in practice. Most constructivists are local constructivists: they hold that some specific variety of normativity is constructed, some particular type of norms, but that others may work in a different way. Rawls, for example, is a constructivist about justice. Kant was a constructivist about duties. T. M. Scanlon’s contractualism has been interpreted as a kind of very local constructivism about right and wrong (Scanlon Reference Scanlon2014; Wallace Reference Wallace2006). I will argue below that Habermas is a constructivist about moral principles, in his sense of the term.

Finally, although Kantian constructivism remains the most influential and well-developed variety, other constructivisms have emerged in recent years, drawing on the work of other philosophers. All these types of metaethical constructivism share some basic insights and claims: that valid moral norms are constructed, under certain constraints, rather than discovered, and that being constructed under these constraints is what gives them their validity; constructed norms as such are neither ‘objective’ like the laws of nature, nor ‘subjective’ like an individual’s whims and desires (though what their ontological status is is an open question). Different varieties of constructivism flesh out these insights in different ways, depending on which philosophical tradition they make use of. Humean (Lenman Reference Lenman2010; Street Reference Street, James and Yonatan2012) and Aristotelian (LeBar Reference LeBar2008) constructivism were the first to emerge, and by today philosophers are at work developing Spinozian (Zuk Reference Zuk2015), Nietzschean (Silk Reference Silk2015), and Hegelian (Laitinen Reference Laitinen, Italo and Luigi2016) constructivisms.

3. Habermas: The Discourse Theory of Morality

Jürgen Habermas famously described his discourse theory of morality as deontological, ‘cognitivist, universalist, and formalist’ (Habermas Reference Habermas1990a: 120–2). Can it also be described as constructivist?

Habermas certainly considers his theory to be reconstructivist. He sees it as a rational reconstruction of really-existing practices of moral argumentation. ‘Moral’ in Habermas’ sense means what is right or obligatory for all subjects, as opposed to ‘ethical’, which concerns what is good for a particular individual or community, in their biographical or social context. Rational reconstruction makes these implicit practices explicit, bringing to light the capacities and procedures we use in coming to agreement about contested moral questions. Although the concrete practices of moral argumentation which the theory reconstructs may be those that exist only in modern societies among subjects with a postconventional moral orientation, Habermas claims that the communicative capacities involved are anthropologically deep-seated, species-wide generic competences (Habermas Reference Habermas1979: 14; 1992: 14).

Moral norms are valid if they meet the conditions of two principles, these principles being idealizations, based on rational reconstruction, of our concrete practices of moral argumentation. First, the discourse principle (D), which holds for all practical discourses, pragmatic and ethical, as well as moral:

Secondly, the universalization principle (U), which holds for moral discourses specifically:

Principle (U) clearly has similarities to the universal law formula of Kant’s categorical imperative (G, 4: 422).Footnote ¹ The key difference between Kant and Habermas, of course, is that the former’s construction procedure (or procedure for testing candidate norms) is monological, and the latter’s is dialogical. It is not enough for the individual moral agent to apply the categorical imperative to their proposed actions, on their own; real moral discourse must take place between interlocutors.

But should we consider Habermas a constructivist? Some of his comments suggest that he would reject the constructivist label. In ‘Moral Consciousness and Communicative Action’, emphasizing again that it does not set up substantive normative orientations, he says that in his theory ‘(p)ractical discourse is a procedure for testing the validity of hypothetical norms, not for producing justified norms. It is this proceduralism that sets discourse ethics apart from other cognitivist, universalist, and formalist ethical theories, and thus from Rawls’ theory of justice as well’ (Habermas Reference Habermas1990a: 122). This suggest that, while Habermas certainly sees the discourse theory of morality as proceduralist, he would resist labelling it as constructivist, like Rawls’ theory of justice – it tests, but does not produce, valid moral norms.

In later texts, however, Habermas clarifies his position. In ‘Remarks on Discourse Ethics’, discussing Rawls’ version of Kantian constructivism, he says the following:

Habermas makes the point that, despite appearances to the contrary, metaethical construction is not voluntaristic. We do not select the construction procedure on a whim, in fact we do not select it at all: it ‘imposes itself on us’. He also emphasizes the theory’s dual aspect. Looked at in one way, its procedures are construction procedures, and thus seem to incline to voluntarism; looked at in another, they are discovery procedures (if moral validity is analogous to theoretical knowledge), and thus seem (‘overhastily’) to incline to a kind of realism. Presumably, the discourse theory of morality’s basis in rational reconstruction is what makes Habermas so confident that the construction procedure is not arbitrarily selected, but rather ‘imposes itself on us’ – he thinks that the procedures of construction or moral discourse, (D) and (U), are rationally reconstructed from the actual argumentative practices of (modern, postconventional) subjects. They thus have a ‘quasi-transcendental’ status in Habermas’ sense of the term (Habermas Reference Habermas2008a).

Looked at in constructivist terms, Habermas is a local, not a global, constructivist. He holds that valid moral norms are constructed through a particular procedure, moral discourse, with a particular set of constraints, principles (D) and (U). It is this procedure under these constraints which gives moral norms their validity. But other normative entities such as ethical values, conceptions of the good for individuals and communities, do not work this way. (D) may apply to them, since it applies to all practical discourses, but (U) does not.Footnote ²

Metaethically, then, the discourse theory of morality should certainly be seen as constructivist, specifically as an example of local constructivism about moral norms. Moral discourse is its construction procedure, and principles (D) and (U) are its constraints. While Habermas’ theory is certainly influenced by Kant, what sets it apart from Kantian constructivism is its dialogical rather than monological nature, and the rationally reconstructed, rather than a priori, status of (D) and (U).

4. Arbitrary or Realist? The Euthyphro Problem for Constructivist Metaethics

One of the most incisive criticisms of constructivism concerns a Euthyphro-like dilemma which it supposedly faces when it tries to justify the constraints placed on the metaethical construction procedure (Wallace 2012: 33–4; Baiasu Reference Baiasu2016: 1186, n. 2). As expressed by Russ Shafer-Landau, the problem is this: all cognitivist versions of constructivism place some constraints on their procedures for constructing valid norms. These may include ‘full information and flawless instrumental reasoning, universalizability of maxims, principle selection from behind a veil of ignorance, etc’. For the discourse theory of morality, these constraints would of course be (D) and (U). The question is, what is the status of these constraints? ‘Either the initial conditions of choice or attitude formation are moralized or they are not. In other words, we are to envision the initial conditions as already incorporating moral constraints, or as operating free of such constraints’ (Shafer-Landau Reference Shafer-Landau2003: 42). The first horn of the dilemma is that the constraints on the metaethical construction procedure are moral; the second horn is that they are non-moral.

Shafer-Landau thinks that the problem with the second horn of the dilemma, where the constraints on the construction procedure are not moral, is that it may lead to the construction of principles which do not at all align with our deepest moral convictions. How serious this problem is depends on whether we see moral philosophy as having a primarily descriptive or a primarily critical role. I will return to this point in the final sections. Mark Timmons maps out a different problem with the second horn of the dilemma. On the assumption that the constraints on the construction procedure are, as he puts it, ‘morally thin’ or ‘relatively morally neutral’, then ‘we get serious indeterminacy … one can only expect some vague and indeterminate moral principles will be selected as a result of the process of moral reasoning so characterized’ (Timmons Reference Timmons2003: 400–1). The worry here is that a construction procedure constrained by these ‘thin’ principles can only construct ‘thin’ norms, norms which are not substantive enough to do much moral work for us.

A more serious implication of the second horn, and perhaps what Shafer-Landau and Timmons are trying to get at, is that it seems to make constructivism arbitrary. If the constraints on the construction procedure do not have moral validity of some kind, then how are constructivists justified in opting for just that set of constraints, rather than any other? Why opt for the categorical imperative rather than the veil of ignorance, if neither of these options is morally charged? The choice of (non-moral) constraints would seem to be arbitrary, and as such constructivism itself would have an arbitrary basis. Arguably, the second horn of the dilemma collapses constructivism into subjectivism: moral agents select a set of constraints, without any moral weight of their own, for the construction procedure they engage in, basing their selection on nothing more than willkür.

The first horn of the dilemma seems a more promising basis for constructivism. And Rawls’ remark that ‘apart from the procedure of constructing the principles of justice, there are no moral facts’, might imply that this is the obvious option. However, its implications seem to be even more damaging:

If the constraints on the construction procedure are moral ones, then where do they come from, or more precisely, where do they get their moral validity from? Not construction, evidently. It looks as if the constraints, in this case, have to be pre-existing moral norms, with a source of validity apart from the construction procedure – in other words, independent moral facts, as understood by realists. As Timmons puts it, they are ‘morally thick’ constraints that involve ‘substantive, non-neutral moral assumptions’ (Timmons Reference Timmons2003: 400–1). If constructivism must rely on pre-existing moral facts to govern its construction procedure, then it seems to collapse into moral realism. It would be, at best, an addendum to realism, holding that some lower-level norms are constructed, and other higher-level, ‘real’ ones, unconstructed – an extremely local constructivism, and of doubtful use.

5. Kantian Constructivism’s Response: The Constitutivist Argument

The first horn of the dilemma, in which the constraints on the metaethical construction procedure are moralized, seems to collapse constructivism into realism. The second horn, in which the constraints are non-moral, seems to collapse it into subjectivism. Faced with this Euthyphro-like problem, most constructivists have opted for the first horn. The challenge which faces them then is to show that the constraints are not pre-existing moral facts, but rather moral entities of some kind whose existence is explicable in constructivist terms. Kantian constructivists, Christine Korsgaard in particular, have responded with what has come to be called the constitutivist argument. This amounts to arguing that the constraints on the construction procedure are constitutive standards of human agency itself, built into the way we as autonomous rational agents operate, in our theoretical as well as practical reasoning. As such, they are indispensable for us, internal to all our actions, and their presence as constraints on the construction procedure does not have to be justified further.

Korsgaard defines the idea of an internal or constitutive standard of something as a standard that is part of that thing being what it is. ‘It belongs to the nature of the object or activity that it both ought to meet, and in a sense is trying to meet, that standard’ (Korsgaard 2008: 7). She illustrates this using the examples of a house and an encyclopedia. The function of a house is to provide shelter, so being waterproof is a constitutive standard for houses; the function of an encyclopedia is to provide information, so it is a constitutive standard for encyclopedias that the information they contain should be true. A constitutive or internal standard for a thing differs from an external standard, which is not connected to that thing being what it is. An external standard for a house might be that it should be beautiful, and an external standard for an encyclopedia might be that it should be light enough to carry around. An ugly house is still a house if it provides shelter from the elements, and a heavy encyclopedia is still an encyclopedia if it conveys truthful information. A structure which does not at all provide shelter, such as a piece of outdoor sculpture, is not a house; a work of fiction written in the form of an encyclopedia, such as one of Borges’ fictional texts, is not an encyclopedia at all. Korsgaard notes that constitutive standards are simultaneously normative and descriptive:

If an object or activity fails to meet its constitutive standards, we may either describe it as a bad, botched example of that object or activity, or as not being an example of that object or activity at all. The dividing line between the two is not always clear. We can call a house with a leaking roof a bad example of a house, or, if it’s in no way waterproof, say that it is not a house at all (but rather a ruin, or a piece of outdoor sculpture). We can call an encyclopedia containing some false or outdated information a bad, unreliable encyclopedia, or if the situation is bad enough, not an encyclopedia at all – in Korsgaard’s example, we may say ‘That’s not an encyclopaedia: it’s just a compendium of nineteenth-century opinion!’ (Korsgaard 2008: 8).

Functional objects such as houses and encyclopedias, in a rather Aristotelian manner, provide the clearest examples of constitutive standards. But the same goes for agency, according to Korsgaard. Certain standards are constitutive of human rational agency: we would not be agents if we did not meet these standards. Korsgaard thinks that both of Kant’s principles of practical reason, the hypothetical and categorical imperatives, are such standards (Korsgaard 2009: 68, 76; G, 4: 414) They are constitutive principles of willing – the activity would not be willing if they did not constitute it. The former makes willing efficacious, the latter makes it autonomous.

So it is not at all question-begging, on Korsgaard’s view, for the hypothetical and categorical imperatives to constrain the metaethical construction procedure. Other constructivists propose other constitutive standards. Yonatan Shemmer, although not strictly making a constitutivist argument, proposes a ‘principle of coherence’ (Shemmer Reference Shemmer, James and Yonatan2012: 164), and Carla Bagnoli proposes a standard of respect for others, aligned to the humanity formula of the categorical imperative (Bagnoli Reference Bagnoli and Carla2013: 176–9; 2016: 1238–40). These constitutive standards are not constructed, but nor are they independent moral facts in the realist sense. They are unavoidable for rational autonomous agents – more than unavoidable, they are internal to agency – so it is no surprise, and no problem, that they appear as constraints in the metaethical construction procedure.

The constitutivist response which Kantian constructivists make to the Euthyphro problem, then, involves claiming that the constraints on the construction procedure (1) are moral, but (2) are standards internal to or constitutive of human agency itself, and so their presence is not question-begging and does not collapse constructivism into realism. As Sorin Baiasu puts it, the constitutivist strategy

It has been objected that this is a circular argument, or an instance of bootstrapping. Kantian constructivists typically respond that any circularity here is inevitable, and virtuous. As O’Neil puts it, if the standards of practical reasoning are fundamental to all human reasoning, ‘then any vindication of these standards is either circular (since it uses those very standards) or a failure (since it is not a vindication in terms of the standards said to be fundamental)’ (O’Neill Reference O’Neill1989: 29).

6. The ‘Shmagency’ Objection

However, the constitutivist argument has failed to satisfy critics of constructivism. One strong line of objections argues that agency (and the standards that constitute it) is not as inescapable as Kantians assume, or that if it is, no metaethical significance attaches to it. David Enoch’s so-called ‘shmagency’ objection makes this case (Enoch Reference Enoch2006, 2011). As Enoch sees it, Korsgaard and others have responded to the sceptical challenge ‘why should I be moral at all?’ with the constitutivist argument ‘because otherwise you won’t be an agent’. But this does not settle the question, since the sceptic can then simply extend their scepticism and ask, ‘but why should I be an agent at all?’ (Enoch 2011: 209). Responding to the version of constitutivism outlined by Korsgaard in Self-Constitution (Korsgaard 2009), that self-constitution is itself a constitutive standard of action, Enoch makes his case in these terms:

On the one hand, Enoch argues that the constitutivist argument will carry no weight for a sceptic who regards terms such as ‘agent’, ‘rational’, etc., as mere bits of folk-psychology. Why should they care if the things they do fall under the description of ‘real actions’ or ‘mere bodily movements’? More importantly for constructivism’s constitutivist strategy, Enoch argues that agency is in fact optional, and that we can choose whether to be agents, who meet whatever constitutive standards of agency the constitutivist specifies, or shmagents, who do not. In terms of the house-building analogy, it is reasonable to say to the builder that, so long as they are engaged in house-building, they must make sure that the structure they are building is waterproof, since this is a constitutive standard of being a house. But it is open to the builder to reply, ‘Very well then, I guess I am not engaging in the project of building a house but rather in the project of building a shmouse, of which these standards aren’t constitutive. So what is it to me how you classify my project?’ (Enoch Reference Enoch2006: 179). Building a house is optional, since one may equally choose to build a shmouse, and the same goes for being an agent – one may equally choose to be a shmagent. If house-building is optional, the constitutive standards of that activity only have normative force for someone who has opted into it; similarly, if agency is optional, then its constitutive standards only have normative force for someone who has chosen to be an agent (Enoch Reference Enoch2006: 186).

The upshot of Enoch’s argument is that agency is in fact as optional as any other kind of activity, and I can choose not to exercise it as I see fit. The implications for constructivism are clear. If the standards internal to agency – the hypothetical and categorical imperatives, according to Korsgaard – are not built into all human activities, then there is no obvious reason why they should appear as constraints on the metaethical construction procedure. If they do, some further explanation for their presence is called for, and we are back at the Euthyphro-type dilemma. The constraints on the metaethical construction procedure would in this case be moral, but contingent, meaning that moral facts antecedent to the construction procedure would be needed to justify them.

The strength of the shmagency objection depends on whether agency does in fact work this way. It is tempting for defenders of constitutivism, especially Kantian constructivists, to simply assert that agency is unavoidable, contrary to Enoch’s claims. Am I not exercising my agency even in choosing not to be an agent? However, Enoch argues that even if agency was unavoidable, nothing morally significant would follow from this. He takes up the analogy of a chess game from David Velleman (Enoch Reference Enoch2006: 185–87; Velleman Reference Velleman2009: 209). Games have constitutive aims, which give rise to normative reasons. If you are not trying to checkmate your opponent, you are not playing chess, and the fact that you are playing chess gives you good reasons to move your knight so as to box in your opponent’s queen, and so on. If some hypothetical group of chess-players could not help but play chess, then it would seem that the constitutive standards of the game would be normative for them. If agency is inescapable for human beings in the same way that playing chess is inescapable for these chess-players, then it seems that the constitutivist account of normativity is vindicated.

Enoch, however, denies that the inescapability of an activity gives any moral weight to that activity’s constitutive standards. ‘In order to have a reason to checkmate your opponent, it seems to me it is not sufficient that you do in fact play chess. Rather, it is also necessary that you have a reason to play chess (or to play the relevant specific game of chess)’ (Enoch Reference Enoch2006: 185). It is not enough, Enoch believes, to claim that some standards are constitutive of agency, and that agency is inescapable for us; for any normative significance to attach to these standards, we also need to have good reasons to be agents, and these the constitutivist cannot provide.

The normative status of the constitutive standards of an inescapable activity is, at the very least, an open question. The mere fact of inescapability does not settle it (Enoch Reference Enoch2006: 189–90). Expanding on Enoch’s point, Christine Bratu and Moritz Dittmeyer use the example of Anna, who is strapped to a machine which forces her to press a red button every ten minutes by pressing her finger down on it. It might well be pointless for Anna to ask what reasons speak in favour of her pressing the button, since the machine is going to force her to do it in any case. But the question is at least meaningful, and ‘because you’re going to do it anyway’ would not be an adequate answer (Bratu and Dittmeyer Reference Bagnoli2016: 1136). Unavoidability does not automatically give rise to normativity. And this presents a serious problem for the constitutivist argument, which aims to ground normativity in the constitutive standards of agency alone: ‘if we need a normative judgment – that we have a reason to be agents rather than shmagents – in order for the constitutive-of-agency strategy to kick in, then the constitutivist strategy cannot give us the whole story of normativity’ (Enoch Reference Enoch2006: 186).

This last point bears elaborating. The constitutivist or constructivist might accept Enoch’s claims but continue to argue that the constitutive standards of agency are the appropriate constraints for the metaethical construction procedure, and to argue further that we should be agents, and thus incorporate those standards into everything we do, including constructing norms. A number of arguments to the effect that it is better to be an agent than a shmagent could be offered. But, as Enoch points out, this response undermines the constitutivist project:

Once again, the constitutivist or constructivist would be forced to rely on antecedent moral norms, contrary to their own commitments (O’Hagan Reference O’Hagan2014: 19).

Enoch’s arguments are not, in my view, fully successful. Constitutivists have defended their position, with some arguing that Enoch’s shmagent is unintelligible, or falls into a performative contradiction in rejecting agency (Velleman Reference Velleman2009; O’Hagan Reference O’Hagan2014; Smith Reference Smith2015), and others concentrating on the issue of the inescapability of agency (Ferrero Reference Ferrero2009; Silverstein Reference Silverstein2012, 2015). The constitutivist approach has in many ways stood up well to the schmagency objection. Nonetheless, it is fair to say that this approach, which accompanies the first horn of the Euthyphro dilemma, has failed to satisfy critics of constructivism. The most promising attempt to classify the constraints on the metaethical construction procedure as moral has run into difficulties. Is there a better alternative?

7. The Second Horn of the Dilemma: The Discourse Theory of Morality

Habermas’ discourse theory of morality can provide an alternative response to the Euthyphro problem of constructivism, which avoids some of the issues besetting Kantian constructivism’s constitutivist argument.

The discourse theory of morality, unlike all strictly Kantian varieties of constructivism, can take the second, apparently unpromising horn of the Euthyphro dilemma: the constraints on the construction procedure are non-moral, in Habermas’ specific sense of the word ‘moral’. The difficulties associated with having unconstructed norms as constraints on the construction procedure, which led Kantians to develop the constitutivist argument, thus dissolve (although other difficulties may well arise). I will argue that the supposedly negative results of casting the constraints on the construction procedure as non-moral would not affect a Habermasian constructivism, properly understood. These ambitious claims can be realized if we look carefully into the status of (D) and (U).

This strategy seems at first to be flatly incorrect: the discourse and universalization principles are surely moral principles – the very fact that Habermas calls them ‘principles’ (not ‘values’, ‘norms’, ‘rules’ or anything else) suggests this. However, remember that (D) and (U) are meant to be rational reconstructions of the concrete communicative practices of human beings, their practices of resolving conflicts and establishing the moral validity of candidate norms. Recall that (D) is meant to be a general principle for all action norms in all practical discourses, not just moral norms in moral ones – as such it is ‘both more and less than a moral principle’, as William Rehg puts it (Reference Rehg1997: 65). As Habermas originally conceived it, the discourse theory of morality was supposed to derive (U) from non-moral premises, either by deduction from facts about the communicative practices of human beings, as described in his formal pragmatics and theory of communicative action, or, in later formations, by abduction from his theory of modernity (Rehg Reference Rehg1997: 65–9; Finlayson Reference Finlayson2000: 320, 331–5).

At the point at which we reconstruct them, then, (D) and (U) are not moral norms, and this allows the discourse theory of morality to opt for the second horn of the Euthyphro dilemma. But in this case what are they? Are they ethical values, pragmatic action norms, or some other kind of normative entity in Habermas’ taxonomy? In my view they should be seen as regulative ideals. As Rehg says, (D) and (U) are ‘idealizations’, ‘having somewhat the nature of regulative ideas (in the Kantian sense) one must rely upon to criticise empirically existing discourses and institutions. As constitutive of real discourse, however, they are also more than regulative ideals’ (1997: 65). In his discussion with Karl-Otto Apel, Habermas makes a point of arguing that while the unavoidable ‘pragmatic presuppositions of argumentation’ – inclusivity, equal distribution of communicative freedoms, truthfulness and absence of external and internal constraints – have a certain normative force, they are not moral. ‘The (weak)-transcendental presuppositions of argumentation differ from moral obligations in that they cannot be systematically violated without destroying the game of argumentation as such. By contrast, we do not have to abandon the moral language game when we violate moral rules’ (Habermas Reference Habermas2008b: 83). Habermas’ intention here is to argue against Apel’s transcendental-pragmatist project of deriving morality as a whole from the presuppositions of argumentation; but in the process, he makes it clear that, considered as constraints on the metaethical construction procedure, they should not be classified as moral.

It is notable that (D) and (U) do not share most of the features which Habermas attributes to moral norms, when he distinguishes them from ethical values and other normative entities. In Between Facts and Norms, he lists four such features. First, moral norms are deontological, not teleological. They ‘obligate their addressees equally and without exception to satisfy generalized behavioural expectations’, while ethical values are only ‘intersubjectively shared preferences’, specific to certain social groups (Habermas 1996: 255). Secondly, moral norms are binary – they are attached to a validity claim which can be answered with a ‘yes’ or a ‘no’, as either valid or invalid. In contrast ethical values, since they express intersubjective preferences for goods, can be gradated – we may value certain goods to a greater or lesser extent. Thirdly, moral norms are absolutely binding, in the sense that their ‘oughtness’ ‘has the absolute sense of an unconditional and universal obligation; “what one ought to do” claims to be equally good for all’ (Habermas 1996: 255). Ethical values, on the other hand, are only relatively binding, telling us what is good for me, or for us, within a certain culture or form of life. Finally, moral norms must not contradict each other, but must form a coherent set. Ethical values, in contrast, can and often do conflict, and an individual or a community need not have a coherent system of values.

The first feature of moral norms, arguably, does apply to (D) and (U) – anyone engaging in moral discourse is obligated to abide by the two principles, which count as rules rather than merely shared preferences. The other three features do not apply, however. (D) and (U) do not raise validity claims to which we can respond with a binary ‘yes’ or ‘no’, but rather describe how a certain type of validity claim, moral ones, can be thematized and criticized. The outcome of the process of moral discourse may be summarized in a binary way – the norm is found to be either valid or invalid – but this is a separate point. Within the context of moral discourse, (D) and (U) certainly have the force of unconditional obligations, but they do not pick out ‘what one ought to do’ in the sense of ‘what is equally good for all’. Instead, they regulate how we arrive at a consensus about what is equally good for all. The issue of coherence with other norms does not arise for (D) and (U). On this basis, I claim that (D) and (U) are not moral in Habermas’ sense of the term. It is worth noting that (D) and (U) do not share the characteristics of ethical values, either.

Does the presence of constraints which are not themselves moral norms, governing the procedure for the construction of moral norms, damage the discourse theory of morality, considered as a constructivist theory? Not if we regard Habermas as a local constructivist, a constructivist about valid moral norms only, using his definitions of moral norms. The validity of all action norms (moral, ethical and pragmatic) is subject to (D), but nowhere does Habermas say that ethical values or pragmatic action norms have to be endorsed in a construction procedure governed by both (D) and (U) in order to be valid. And nowhere does he indicate that he is a constructivist about all forms of normative validity.

8. An Alternative to the Constitutivist Argument

So, the discourse theory of morality, unlike more strictly Kantian varieties of constructivism, can take the second horn of the Euthyphro dilemma, and hold that the constraints on the construction procedure are non-moral (in Habermas’ sense of ‘moral’). Thus, it has no need to adopt the constitutivist strategy, and avoids the problems of the first horn. Various worries were expressed about this option, and if the discourse theory of morality is to succeed as a variety of constructivism, these worries must be addressed. In each case, I will argue that Habermas’ theory in fact fits better with the spirit of metaethical constructivism than other varieties.

Shafer-Landau worried that, if the constraints on the construction procedure are non-moral, then the principles they generate will not align with our moral convictions. As he puts it, ‘there is no reason to expect that the principles that emerge from such a construction process will capture our deepest ethical convictions, or respect the various platitudes that fix our understanding of ethical concepts’ (Shafer-Landau Reference Shafer-Landau2003: 42). This is perhaps not as serious a problem as Shafer-Landau thinks it is. Our deepest ethical convictions are by and large the products of time, culture and tradition, and they may well be wrong about many things and in need of correction. If metaethical constructivism gives rise to principles that challenge our most deeply held moral convictions, then so much the worse for (at least some of) our convictions. Habermas’ discourse theory of morality is, after all, part of the overall project of a critical theory of society. If, in the spirit of this project, we see his moral philosophy as critical, aiming to diagnose and rectify social pathologies, rather than descriptive, aiming to give an account of our moral convictions as they stand, then Shafer-Landau’s worry becomes a point in favour of the Habermasian version of constructivism.

One of the most insightful discussions of the implications of classifying metaethical construction procedures as non-moral comes from R. Jay Wallace. As he sees it, construction procedures must have enough normative weight for individuals so that they can use them to correct their mistaken moral beliefs. This is an essential component of any moral theory and fits well with the critical aim of the discourse theory of morality in particular. Supposing (as constructivists must) that the construction procedures are not antecedently moral, this normative weight must come from ‘a dialectical account of the normativity of the rational procedures’:

It is the fact that individuals are, as matter of fact, committed to the principles which appear as constraints on the metaethical construction procedure which gives the principles normative weight, despite the fact that they are not antecedently moral. For this to work, these commitments, according to Wallace, must be ‘actual normative attitudes’ and not ‘true normative commitments’. By this latter term he means idealizations or rationalizations, ‘the attitudes we would arrive at if we subjected our actual normative views to this kind of procedural correction’ (Wallace 2012: 33). The latter case would be circular – Wallace’s assumption is that our ‘true normative commitments’ are arrived at by testing our ‘actual normative attitudes’ using some procedure such as the categorical imperative, the veil of ignorance or (D) and (U). If it is these rarefied ‘true normative commitments’ which carry weight for us, the principles would be self-validating. Only unreflexive ‘actual normative attitudes’ will do as non-moral constraints on the construction procedure.

However, Wallace identifies a problem with this. As he puts it,

The problem is that, if only actual normative attitudes will do, then only fairly simple principles can act as non-moral constraints on the construction procedure. Wallace gives the hypothetical imperative and the principle of non-contradiction as examples. The more complex principles favoured by more ambitious, Kantian varieties of constructivism – such as the categorical imperative in Korsgaard’s version – would not satisfy this condition. Most people are not walking around with the universal law formula at the forefront of their minds. This, it seems, is a challenge to any variety of constructivism which classifies its metaethical construction procedures as non-moral, and which aspires to be critical rather than descriptive.

How does the discourse theory of morality fare in this respect? At first glance, (D) and (U) would seem to be paradigmatic cases of idealized, reflexive, ‘true normative commitments’, and thus subject to Wallace’s criticism. However, the crucial point is that Wallace considers actual normative commitments which the individual acknowledges either implicitly or explicitly to be suitable as non-moral constraints on the construction procedure. ‘(T)he procedural standards can do the work the constructivist allots to them only because they are standards that agents can be assumed already to acknowledge, implicitly or explicitly. Their authority, on this account, can be traced to the fact that they are standards that the agent actually accepts’ (Wallace 2012: 35). For Wallace, ‘actual acceptance’ and ‘implicitness’ can evidently go together. Habermasian constructivists can argue that postconventionally moral modern language-users do indeed endorse (D) and (U) implicitly, and that the discourse and universalization principles themselves are simply explications. The strength of the Habermasian variant of constructivism would then depend on how robust this claim is, and if it can meet Wallace’s ‘stringent condition’.

It is certainly true that the average person is not walking around with (D) and (U) at the forefront of their mind, any more than the categorical imperative. The method of rational reconstruction which Habermas uses to arrive at the discourse theory of morality, however, aims at making implicit procedures and capacities explicit (Pedersen Reference Pedersen2008: 461–5). According to Habermas, we draw on deep-rooted, species-wide communicative competences when coming to agreement on the validity of moral norms. The particular way those of us who are postconventionally moral, modern language-users do so is only implicitly known to us, until rational reconstruction makes it explicit in the form of (D) and (U). In Wallace’s terms, the discourse and universalization principles are indeed ‘true normative commitments’, idealized and rationalized; but they are at the same time the explicated versions of the ‘actual normative attitudes’ we hold implicitly. As such they are suitable to act as the non-moral constraints on the metaethical construction procedure. These constraints are not self-validating, as Wallace worries. (D) and (U) are not generated using (D) and (U), but rather through the separate process of rational reconstruction.

The most serious concern about taking the second horn of the Euthyphro dilemma was that if the constraints on the construction procedure are non-moral, they must be arbitrarily selected, and thus constructivism itself must be arbitrary. By implication, any constructed norms would themselves have an arbitrary status. In Baiasu’s words, ‘The fundamental objection is that if the construction is not normatively constrained, then norms are arbitrary; if it is normatively constrained, then it relies on some normative constraints which are not constructed’ (Baiasu Reference Baiasu2016: 1186, n. 2). After all, if the constraints are not themselves morally loaded, why select one particular set of constraints, such as (D) and (U), and not another, like the hypothetical and categorical imperatives, or the veil of ignorance?

However, Habermas’ principles (D) and (U) are not at all arbitrary: as he says, they ‘impose themselves on us’. As we have just seen, they are rationally reconstructed from concrete practices of discourse and conflict resolution, which are themselves connected to deep-seated, species-wide communicative capacities. Even if Habermas’ original ambition of deriving (D) and (U) from formal pragmatics alone does not succeed, as Finlayson argues, then the principles still stand as rational reconstructions of the processes of conflict resolution engaged in by postconventionally moral language-users under conditions of modernity. Now, in a sense our status as postconventionally moral modern language-users is contingent. Habermas is no Hegelian, and his philosophy of history does not hold that the cunning of reason inevitably leads to a fully rational society; diversions, regressions and alternative paths of development are possible. And yet, here we are. This aspect of Habermasian constructivism fits well with one of the fundamental insights of constructivism, namely that morality is made by human beings, and that even when it claims universal validity, it claims it from a particular point in time and space. In any event, (D) and (U) are not selected at random, and their status as constraints on the metaethical construction procedure is not arbitrary.

In saying this, am I falling back into a type of constitutivist argument, similar to the one used by Kantian constructivists, which would be vulnerable to the same criticisms? Am I making an argument to the effect that the communicative practices from which (D) and (U) are reconstructed are unavoidable for human beings, constitutive of some inescapable facts about them? No: while these practices of conflict resolution and moral discourse may be widespread among subjects with a postconventional moral orientation, in modern societies, and the communicative capacities they draw on may be deep-seated and species-wide in human beings, Habermas does not claim that they are inevitable and inescapable, as Kantian constructivists making the constitutivist argument claim agency is inevitable and inescapable. The communicative capacities in question may be as close to ‘universal’ for human beings as Habermas’ postmetaphysical philosophical framework allows, but the concrete practices – and it is these which are the fodder for rational reconstruction – are the products of historical development. Understood in this way, the claim that (D) and (U) ‘impose themselves on us’ is strong enough to cancel out the threat of arbitrariness, but weak enough that it does not replicate the constitutivist argument, with all that entails.Footnote ³