1 Introduction
Amongst metaethicists and Kant scholars, there continues to be a lively debate about whether Kant is best understood as a constructivist, a realist or something in between.Footnote 1 One thing that is accepted by all sides, however, is that if Kant were a constructivist, he would be what is called a ‘Kantian constructivist’, i.e. he would believe that moral truths can be derived by reflecting on the constitutive features of practical agency. To this end, it has become common to understand Kant’s constructivism by means of an analogy to the theoretical sphere: just as logic is constitutive of theoretical thinking, so too the Categorical Imperative is constitutive of practical thinking. Kant commentators like Korsgaard, for example, argue that the Categorical Imperative is just a ‘principle of the logic of practical deliberation’ (2008: 321), Herman has described Kantian arguments as an attempt to ‘derive a substantive moral result from a purely formal constraint of rational willing’ (1993: 230), and Reath (Reference Reath2013: 201) has said that the Categorical Imperative is a formal principle. When it comes time to assess whether Kant was a constructivist or a realist, this interpretation is then accepted by both sides. For example Stern, who argues that Kant is a realist, says that on a constructivist reading the Categorical Imperative should be compared to a law of logic (2013: 23) and Sensen, who argues that Kant is somewhere in between a realist and a constructivist, compares the Categorical Imperative to the principle of non-contradiction (2013: 64–5). And, of course, that Kant believed that morality was a constitutive feature of practical thinking is picked up and legitimated by other metaethicists whenever they refer to this position as ‘Kantian constructivism’ (e.g. Street Reference Street2010: 369–70, Reference Street2012: 41; Bagnoli Reference Bagnoli2014; Fitzpatrick Reference Fitzpatrick2013: 42; LeBar Reference Lebar2013: 184). In this paper, I will argue that this consensus is mistaken and that Kant should not be understood as a ‘Kantian constructivist’. My first move in this direction is to relabel ‘Kantian constructivism’, and I will therefore call the position that holds morality to be derived from constitutive features of our agency austere constructivism.
If austere constructivism is true and morality is constitutive of our rational agency, then it looks like morality might be part of the definition of rational agency. In other words, there would be an analytic connection between rational agency and morality. But throughout his life, Kant explicitly denied that there is an analytic connection between morality and rationality. Thus in the Groundwork, he says:
That this practical rule [the principle of autonomy] is an imperative, that is, that the will of every rational being is necessarily bound to it as a condition, cannot be proved by mere analysis of the concepts to be found in it, because it is a synthetic proposition … (G, 4: 440. See also G, 4: 420, 444, 447; CPrR, 5: 31, 46; MM, 6: 248; Rel, 6: 7)Footnote 2
And later, in the Religion, Kant gives his most striking and notorious denial:
The most rational being of this world … might apply the most rational reflection to these objects [the objects of inclination] – about what concerns their greatest sum as well as the means for attaining the goal determined through them – without thereby even suspecting the possibility of such a thing as the absolutely imperative moral law … (Rel, 6: 26)
As striking as these passages are, however, it is open to an austere constructivist simply to deny that constitutivity and analyticity are the same thing. Constitutivity, as we will see, has to do with what is entailed from an agent’s point of view and may have little to do with what, from an outsider’s point of view, can be gleaned from the meaning of agency. One of the central aims of this paper will be to show that constitutivity and analyticity are indeed connected. Analytic judgements, I will argue, are just judgements which follow from rules constitutive of a kind of thinking, either practical or theoretical. It will follow that austere constructivism is incompatible with Kant’s belief that morality and rational agency are connected synthetically.
Although my central objection to austere constructivism as an interpretation of Kant has to do with the analytic–synthetic distinction (ASD), austere constructivism will also touch down on Kant’s texts at other points. In particular, it will bear on questions about the role of formal principles in Kant’s ethics, and part of my task will be to show that here too Kant’s positions are incompatible with austere constructivism.
The argument that Kant was not an austere constructivist will be conducted in four sections. In the first section, I will briefly explain austere constructivism and compare it to Humean constructivism and mathematical logicism. In the second section, I will consider direct evidence from the preface to the Groundwork and apparent counter-evidence from the Critique of Practical Reason. In the third section, I will attempt to explain Kant’s decision to apply ASD to the practical sphere, and I will show that it is incompatible with austere constructivism. In the fourth section, I will briefly sketch what an authentically Kantian constructivism might look like and how it is different from austere and Humean versions.
2 Constructivism and Logicism
Although there are many different characterizations of constructivism, here I will follow Street’s ‘practical standpoint characterization’ (Street Reference Street2010: 367).Footnote 3 According to Street’s constructivism, to say a normative claim is true is not to make a claim about some mind-independent reality. Instead, a normative claim is true if it is ‘entailed from within a practical point of view’ (ibid.). While the theoretical point of view asks questions about what is the case, the practical point of view asks questions about what is to be done. Answers to these questions purport to be ‘solutions to practical problems’.Footnote 4 The result of a practical inquiry is a plan of action rather than a description of how the world is. For a constructivist, a normative claim is true because it is a correct solution to a practical problem, not because it is an accurate portrayal of the world.
As Street explains, a normative claim is entailed within a practical point of view if it is constitutive of features of the agent’s practical problem (Street Reference Street2010: 367). In other words, a feature of a solution to a practical problem is entailed from within an agent’s practical point of view if that feature is already contained in the statement of the problem. What it is for a normative claim to be constitutive of practical agency can be made clear by comparing it to theoretical reasoning. In the case of theoretical reasoning, the rules of logic comprise the norms constitutive of the activity of thinking because logic is so necessary for thinking that failing to be governed by the norms of logic is a failure to think at all. As many have put it, such a failure is not a form of thinking badly, it is not thinking at all. This does not mean, of course, that a single affirmation of the consequent overthrows one’s status as a thinker. But it does mean that someone who refused to have her thinking ‘assessed’Footnote 5 by modus ponens would indeed fail to be a thinker. Returning to the practical sphere, a constructivist can claim that an agent who fails to comply with a normative truth (i.e. one that is constitutive of her practical point of view) is like an epistemic agent who fails to comply with the laws of logic. If we take means-ends reasoning as one of these constitutive norms of agency, the parallel to theoretical reason tells us that an agent who refuses to have her actions assessed by means-ends reasoning is failing to be an agent at all.
All constructivists agree on what it means for a normative claim to be true – it is true if is constitutive of a practical point of view. They might even agree that some normative claims, perhaps generic claims about means-ends rationality, are constitutive of any practical point of view. The difference between Humean and austere constructivists, however, is how they classify moral claims. An austere constructivist holds that moral claims, like generic claims about means-ends rationality, are constitutive of the practical point of view as such. In other words, an austere constructivist believes that moral truths can be derived simply by reflecting on the practical point of view common to all agents. A Humean constructivist, on the other hand, believes that moral claims can be derived from particular practical points of view, but the practical point of view as such must be supplemented with the contingent values of particular agents.
To explain this distinction, Street asks us to imagine an ‘ideally coherent Caligula’ (Street Reference Street2010: 371) who enjoys and values torturing others. Austere constructivists claim that even this Caligula has reason to abandon his ways. While Caligula may himself fail to see these reasons, sufficiently careful reflection on the norms of practical deliberation as such ought to reveal, even to someone with the desires of Caligula, that wanton cruelty is incompatible with those norms. A Humean constructivist denies that Caligula would have such reasons. Instead, they believe that nothing from within Caligula’s normative point of view entails that torturing others is wrong and therefore Caligula can have no reason to stop torturing.
Each theory faces its own difficulties. According to Humean constructivism, moral truths end up being relative to an agent’s contingent desires. Since wanton cruelty is entailed from within Caligula’s practical point of view, ‘I ought to wantonly torture’ will come out as a true normative claim. Which claims turn out to be true or false depends on historical accidents.Footnote 6 Austere constructivism, on the other hand, does not lack ambition. It attempts to derive a lot from a very little and has been criticized again and again for its failure to make good on its ambitious claims.Footnote 7 As Street puts it, Humean constructivists ‘deny that the rabbit of substantive reasons can be pulled out of a formalist hat’ (Street Reference Street2010: 370). I will argue that one of the benefits of looking back to the historical Kant is that Kant’s position does not immediately fall prey to either of these two criticisms and the possibility is raised that it ultimately will fall prey to neither.
Humean and austere constructivists differ on how much fecundity they accord to the constitutive laws of practical reasoning. Austere constructivists believe that substantive moral truths can be derived from the merely formal rules of practical thinking whereas Humeans deny that such substantive results can be derived from such premises. The debate can be helpfully compared to the early twentieth-century debate between logicists and anti-logicists. Following MacFarlane Reference MacFarlane2002, we can understand the debate as one over the fecundity of the constitutive laws of thought. Both sides identified logic with the norms constitutive of thinking, but logicists like Frege (MacFarlane Reference MacFarlane2002: 38–40) held that these norms could have substantive (i.e. mathematical) implications. Drawing on this understanding of the dispute between logicist and anti-logicists, we might say that austere constructivists are the logicists of the practical sphere since they hold that substantive ethical claims can be reduced to claims about the ‘logic of practical deliberation’ (Korsgaard Reference Korsgaard2008: 321). Humeans, on the other hand, are the anti-logicists of the practical sphere since they hold that substantive practical claims cannot be derived from constitutive norms of practical reasoning.
Since all sides hold that Kant is an austere constructivist, then all sides must hold that Kant is a practical logicist. This is the first indication that all is not well with austere constructivism as an interpretation of Kant. In the theoretical sphere, Kant is an unswerving anti-logicist. In the Critique of Pure Reason, he insists that general logic (which he has previously identified with the norms constitutive of thinking (A52/B76; JL, 9: 12) is merely formal and has no substantive implications:
But since the mere form of cognition, however well it may agree with logical laws, is far from sufficing to constitute the material (objective) truth of the cognition, nobody can dare to judge of objects and to assert anything about them merely with logic without having drawn on antecedently well-founded information about them from outside of logic … For since it [general logic] teaches us nothing at all about the content of cognition, but only the formal conditions of agreement with the understanding, which are entirely indifferent with regard to the objects, the effrontery of using it as a tool (organon) for an expansion and extensions of its information, or at least the pretension of so doing, comes down to nothing but idle chatter. (A60–1/B85–6; emphasis added)
If the consensus is right about Kant, then Kant was an avowed anti-logicist in the theoretical sphere but a logicist in the practical sphere. This is possible, of course. Perhaps theoretical reason and practical reason are different in ways that would explain how the constitutive norms of one activity can derive substantive results but the constitutive norms of the other cannot. But especially given Kant’s propensities to draw frequent analogies between the two spheres and his belief that ‘there can, in the end, be only one and the same reason, which must be distinguished merely in its application’ (G, 4: 391), this striking dis-analogy can at least serve as motivation to bring the conventional wisdom of Kant’s practical logicism into question.
3 Evidence from the Groundwork and the Critique of Practical Reason
In this and the next section, I will argue that Kant is a thoroughgoing anti-logicist about both theoretical and practical reason – the constitutive rules that govern each of these domains are incapable of having substantive implications. In particular, the rules constitutive of practical reason describe the activity of our practical thought rather than the objects of our practical thought. My argument here will be primarily exegetical. I will make only minimal attempts to show why Kant thought the way he did or why he was right to think the way he did. I will briefly say something about the advantages of Kant’s kind of constructivism in section 5.
The most direct evidence for Kant’s anti-logicism comes from the Groundwork.Footnote 8 In the preface to the Groundwork, Kant divides up rational cognition along three axes: practical/theoretical, pure/empirical, and material/formal. He classifies the metaphysics of morals as material (rather than formal), practical (rather than theoretical) and pure (rather than empirical). As practical knowledge, a metaphysics of morals has to do with the laws of freedom rather than the laws of nature and as pure it does not rest on grounds taken from experience. But most importantly for my purposes, Kant claims that a metaphysics of morals is material cognition rather than formal cognition.
Kant explains that formal cognition is occupied only with ‘the form of the understanding and of reason itself and with the universal rules of thinking in general, without distinction of objects’ (G, 4: 387), whereas material cognition is characterized as ‘concerned with some object’ (ibid.). Kant here references his discussion of general logic in the Critique of Pure Reason where, as we have seen, he claims that formal knowledge produced by logic ‘teaches us nothing at all about the content of cognition, but only the formal conditions of agreement with the understanding’ (A60/B85). He insists that a metaphysics of morals, i.e. everything for which the Groundwork is a groundwork, is material not formal cognition. In other words, the topic of the Groundwork is not formal analyses of practical reasoning, but rather a material investigation into practical reasoning. In words as plain as can be expected, Kant introduces the possibility that he is engaging in a sort of practical analogue of logic and rejects it precisely on the grounds that such an analogue would be merely formal. Whatever else the Groundwork is, it is simply not an attempt to derive contentful ethical conclusions from merely formal practical thinking.
Against this, one might think to place Kant’s frequent assertion in the Critique of Practical Reason that it is possible for the will to be determined by merely a ‘formal practical principle’. Indeed, Kant says that when the will is so determined it is determined by morality. The following are typical of the many times Kant makes these kind of remarks:
and thus either there is no higher faculty of desire at all or else pure reason must be practical of itself and alone, that is, it must be able to determine the will by the mere form of a practical rule without presupposing any feeling and hence without any representation of the agreeable or disagreeable as the matter of the faculty of desire, which is always an empirical condition of principles. (CPrR, 5: 24)
If a rational being is to think of his maxims as practical universal laws, he can think of them only as principles that contain the determining ground of the will not by their matter but only by their form. (CPrR, 5: 27)
These passages express a central Kantian thought. There are two possible determining grounds of the will. If the will is determined by the lower faculty of desire or empirically, it is said to be determined by the object of the will. In this case, we choose the object because of something about the object that is attractive to us. If, on the other hand, the will is determined by the higher faculty of desire, it is said to be determined by the form of the will. In this case, formal properties of our will are the sole considerations that make us choose the object. It is a central tenet of Kant’s moral theory that (1) such a formal determination of the will is possible and (2) the will is so determined precisely when it allows morality to determine it. But if merely formal principles of the will can determine the will to act morally, does not this mean that morality derives from merely constitutive principles of willing as such?
I think that close attention to these passages reveals the reverse message. Consider the first passage. What Kant actually says is that merely formal principles must be able to determine the will if it is possible for there to be a higher faculty of desire. Our faculty of desire is our ability to act practically (i.e. to bring about an object by means of representing them as desirable: CPrR, 5: 9 n., MM, 6: 211) and the higher faculty of desire is the capacity to act practically according to your own laws. Similarly, in the second passage, he says that merely formal principles must be able to determine the will if there are practical universal laws. In the Critique of Practical Reason, universal laws are identified with objective laws (CPrR, 5: 19). In other words, the form determines the maxim only when the will is governed by objective laws. Formal principles govern not all of practical reasoning but practical reasoning that attains to universal laws – i.e. practical reasoning that is objective. In each of these cases, Kant does not assert that formal principles of practical thinking are capable of determining the will. Instead, he argues that, when we practically represent an object, then the formal principles of that activity are capable of determining the will. In other words, morality is said to be constitutive not of the faculty of practical reasoning as such, but constitutive of practical reasoning that is objective. I will return to this in the fifth section.Footnote 9
4 The Application of ASD to Imperatives
In this section, I would like to consider Kant’s decision to apply ASD to imperatives, and I will argue that this decision can only be explained if we assume that Kant was not an austere constructivist. Especially in the Groundwork, but also in later works, Kant claims that hypothetical imperatives are analytic and the Categorical Imperative is synthetic. He says, for example, ‘if only it were as easy to give a determinate concept of happiness, imperatives of prudence would agree entirely with those of skill and would be just as analytic’ (G, 4: 417, my emphasis; cf. 4: 420). The clear implication of this is that imperatives of skill, at least, are analytic. And the case for the Categorical Imperative being synthetic is even more straightforward: ‘It [the Categorical Imperative] is an a priori synthetic practical proposition’ (G, 4: 420; cf. 4: 444; MM, 6: 255; Rel, 6: 6–7, 11).
But what can Kant mean by this? He tells us that a theoretical judgement is analytic if the subject contains the predicate (A6/B10), but how does this metaphor of containment apply in the case of imperatives? Kant commentators have also talked about a characterization of analyticity according to which an analytic judgement is one whose negation is a contradiction. But how do you negate an imperative and what does it mean for an imperative to be contradicted? Since Kant does not tell us how to answer any of these questions, commentators have been puzzled by Kant’s application of ASD to the practical sphere. Some have thought that Kant was speaking metaphorically, others have thought Kant was confused, and still others have attempted to ‘fill up the gaps’ and explain how Kantian principles can justify such an application.Footnote 10 In this section, I will argue that the best explanation for Kant’s application assumes that he is not a practical logicist.
Amongst Kant commentators, it used to be accepted that the characterization of ASD in terms of contradiction was superior to that in terms of ‘containment’ because it applied to more cases and did not rely upon an unexplained metaphor.Footnote 11 Recently, Anderson (Reference Anderson2004, Reference Anderson2005) has challenged this consensus by showing that the containment characterization is a self-standing characterization that is not equivalent to the contradiction characterization.Footnote 12 If we accept these results, as I think we must, then for each usage of ASD, we must figure out whether Kant was thinking about the containment or contradiction characterization.Footnote 13 I will argue that no sense can be made of parts of an imperative containing other parts of imperative, and I will conclude that if we want to understand how imperatives can be analytic or synthetic we must turn to the contradiction characterization.
My argument will proceed in three stages. In the first stage, I will attempt to use the containment characterization to extend ASD to the practical sphere, and I will conclude that containment cannot be extended in this way. In the second stage, I will consider the contradiction characterization and argue that there is a natural extension of it that involves the idea of constitutivity. Finally, in the last stage of my argument, I will show that a consequence of this extension of ASD into the practical sphere is that the rules constitutive of practical thinking must be formally empty.
Containment Analyticity?
I begin by trying to apply the containment characterization of analyticity to hypothetical imperatives. Consider the following hypothetical imperative: ‘if I want to bisect a line, I must make two intersecting arcs’ (G, 4: 417). There does seem to be a relation between two concepts in this hypothetical imperative, namely ‘bisected lines’ and ‘made with two intersecting arcs’, and we could perhaps turn it into the following categorical judgement: all bisected lines are made with intersecting arcs. But, as Kant emphasizes, these two concepts are synthetically related (ibid.). The containment characterization has to do with how the parts of a judgement, i.e. the concepts, are related to each other and the only two concepts within this judgement are synthetically related.
But perhaps when it comes to imperatives Kant is talking not about the relationship between the parts of the judgment but a relationship between our attitudes towards the parts of the judgement. This is already a serious departure from the original containment characterization, which is about parts of concepts and not our attitudes towards them. But Kant himself seems to suggest something along these lines when he distinguishes the synthetic relationship that holds between means and ends and the analytic relationship that holds between our willing of ends and our willing of means:
That in order to divide a line into two equal parts on a sure principle I must make two intersecting arcs from its ends, mathematics admittedly teaches only by synthetic propositions; but when I know that only by such an action can the proposed effect take place, then it is an analytic proposition that if I fully will the effect I also will the action requisite to it; for, it is one and the same thing to represent something as an effect possible by me in a certain way and to represent myself as acting in this way with respect to it. (G, 4: 417)
This passage might suggest that the containment relation holds between my willing of the end and my willing of the means. Intuitively, when we commit to an end, it does seem like willing the means is already contained within that commitment. Given that I already believe that there is no way to bisect a line without constructing intersecting arcs, I just cannot will the former without thereby willing the latter.
But of course I might akratically will the end without willing the means. If there is a relationship between the willing of the end and the willing of the means, it must be normative. So perhaps the containment relationship attains not between actual willings but between the willings of ideal agents. At last, we have a potential definition of analyticity using the containment definition:
Practical Analytic-Containment (PAC): an imperative is analytic if for an ideally rational agent the willing of the means is contained in the willing of the end.
Practical Synthetic-Containment (PSC): an imperative is synthetic if for an ideally rational agent the willing of the means is not contained in the willing of the end.
According to PAC, hypothetical imperatives are analytic because when an ideal agent wills that she bisect a line, a secondary willing that she make two intersecting arcs is already contained therein. Similarly, the Categorical Imperative is synthetic because it is contained in no antecedent willing of even an ideally rational agent. PAC-PSC has at least two major differences from its theoretical cousin. First, it operates on willings rather than concepts. Second, it applies to ideal agents rather than actual agents. While these are serious departures, I think that PAC-PSC remains close enough to its theoretical cousin that so understood it might make sense of Kant’s decision to apply ASD to imperatives.
Unfortunately, PAC-PSC cannot be a valid interpretation of ASD. In the original containment characterization, the metaphor of containment played a crucial role. According to Kant’s understanding of containment, concepts can be analysed into sub-concepts or ‘marks’ (JL, 9: 95) and a judgement is analytic if the set of subject-term marks literally contain the set of predicate-term marks. Thus if ‘triangle’ is analysed into ‘three-sided’ and ‘figure’, then the judgement triangles have three sides is analytic because the predicate term is literally one of the marks of the subject term. As Kant puts it bodies are extended is analytic because ‘To everything x, to which the concept of body (a+b) belongs, belongs also extension (b)’ (JL, 9: 111).Footnote 14
But how can we apply this analysis to the activity of willing? Willings unlike concepts do not seem to be analysable into parts, and if they were it does not seem like the means are literally a part of willing the ends, as if willing the ends is willing the means plus something else. Instead, it seems like what we mean when we say that willing the means is contained in willing the ends is that willing the means is necessarily involved in willing the ends. The containment seems to amount to little more than a claim that it is impossible to do one without the other (provided you are a fully rational agent). But if the relationship between willing the ends and willing the means is one of simple necessity for ideally rational agents, then it becomes unclear why categorical imperatives are not also contained in the willing of our ends. After all, if Kant is right about the connection between agency and morality, then it is impossible for a rational agent to will any end without also willing the Categorical Imperative. Once a perfectly rational agent has set herself an end it is just as necessary that she abide by morality as it is that she will the means.
So what went wrong? Since willings do not have parts, we were forced to reduce analyticity to necessity. But while it is true that an ideally rational agent will necessarily will the means to her ends, understanding analyticity as necessity runs afoul of one of the central ambitions of ASD. As we will see, one of the points of ASD is to distinguish two different types of necessity – the analytic a priori from the synthetic a priori – and understanding analyticity as necessity destroys any hope of making this distinction. In what follows, I will argue that the contradiction characterization fares far better in distinguishing two different types of practical necessity.
Contradiction Analyticity
According to the contradiction characterization, a judgement is analytic if its negation is a contradiction (A151/B190. Cf. B12; P, 4: 266–7; OAD, 8: 229, 230, 231), and it is synthetic otherwise. Of course this definition is only precise if we can say precisely when the negation of a judgement is a contradiction. But when a sentence is a contradiction is a question that partly falls within formal logic. It is no surprise that contemporary philosophers have accordingly defined analyticity in terms of formal logic plus synonymy. Van Cleve expresses this tradition clearly when he writes: ‘A is analytic iff from its negation ~A a formal contradiction may be derived, using in the derivation only laws of logic and substitutions authorized by definitions’ (Van Cleve Reference Van Cleve1999: 20). Provided that we are given a perfect dictionary, then it is just a matter of applying logical manipulations in order to check whether the negation of the sentence leads to a formal contradiction.
According to the contradiction characterization, therefore, there is an intimate connection between analytic judgements and general logic: modulo substitutions of synonymy, an analytic truth is just one whose truth follows from the laws of general logic. Of course the synonymy condition is the one that has received the most attention, but it is not clear that Kant would have agreed that it is the most important. He certainly would have accepted that formal tautologies containing no substitutions of synonymy are analytic (B17; JL, 9: 111). In any case, there is a tight connection between analytic judgements and judgements whose truth is known by logic. In section 2, we followed MacFarlane in noting that for Kant the laws of logic are just the laws that are constitutive of thought. Simply inserting constitutivity into Van Cleve’s contradiction characterization yields:
C-Analytic: P is analytic iff from its negation ~P a formal contradiction may be derived, using in the derivation only laws constitutive of thinking as such and substitutions authorized by definitions.
There seems to be no good reason to keep this as a negation. So equivalently:
C-Analytic*: P is analytic iff P can be derived using only laws constitutive of thinking as such and substitutions of synonymy.
And correspondingly,
C-Synthetic*: P is synthetic iff it is not c-analytic*.
Notice that I have done nothing but substitute Kant’s definition of general logic into the canonical definition of analyticity and syntheticity.
The benefit of translating ASD into statements about constitutivity is that the distinction can now be naturally carried over into the practical sphere. Since agency has laws constitutive of its performance just as thinking does, we could say that a practical claim is analytic if it follows from rules constitutive of practical reasoning.
P-Analytic*: A practical claim Q is analytic iff Q can be derived using only laws constitutive of practical reasoning as such and substitutions of synonymy.
P-Synthetic*: A practical claim Q is synthetic iff Q cannot be derived using only laws constitutive of practical reasoning as such and substitutions of synonymy.
Substitutions of synonymy are not as important in the practical sphere as they are in the theoretical sphere since it is rarely the case that a key issue in practical deliberation depends upon understanding the meaning of the words. So a claim is practically analytic if it can be derived using only laws constitutive of practical reasoning as such.
As we have seen, Kant says that hypothetical imperatives are analytic and the Categorical Imperative is synthetic. Applying the practical understanding of ASD [p-ASD] means that hypothetical imperatives can be derived using only laws constitutive of practical reasoning as such and substitutions of synonymy, and categorical imperatives cannot be so derived. But if the Categorical Imperative cannot be derived using only laws constitutive of practical reasoning as such, then austere constructivists are wrong to attribute such a position to Kant. If we are right to understand p-ASD in terms of constitutivity then Kant’s claim that the Categorical Imperative is synthetic simply rules out an austere constructivist reading of Kant.
As illustrations of this theory, consider how it would apply to a hypothetical imperative and a categorical imperative:
Lying-H: ‘I ought not to lie if I will to keep my reputation.’ (G, 4: 441)
Lying-C: ‘I ought not to lie even though it would not bring me the least discredit.’ (G, 4: 441)
To say that Lying-H is p-analytic is to say that it can be derived from laws constitutive of practical reasoning as such and substitutions of synonymy, and to say that Lying-C is p-synthetic is to say that it does not so follow. But it might well be wondered how Lying-H is going to follow from considerations of our agency as such. Surely, for example, at least some empirical premises about the relationship between lying and losing your reputation must be supplied.
Kant worries about this earlier in the Groundwork. After distinguishing between hypothetical and categorical imperatives, he raises the question of how these imperatives are possible. Famously, he claims that the possibility of hypothetical imperatives is not difficult to explain but the possibility of categorical imperatives is quite difficult to explain:
Whoever wills the end also wills (insofar as reason has decisive influence on his actions) the indispensably necessary means to it that are within his power. This proposition is, as regards the volition, analytic … (synthetic propositions no doubt belong to determining the means themselves to a purpose intended, but they do not have to do with the ground for actualizing the act of will but for actualizing the object). (G, 4: 417; emphasis added)
It is important to note that Kant explicitly acknowledges that synthetic propositions are necessary for determining the means to our ends. As expected, Lying-H depends upon some synthetic empirical propositions about the relationship between lying and our reputation. Nevertheless, Kant insists that these synthetic propositions do not concern the ‘ground for actualizing the act of will’ but instead concern the ground ‘for actualizing the object’. Kant distinguishes here between two kinds of relations. On the one hand there is the theoretical judgement that connects lying with a ruined reputation. This judgement is empirical and synthetic and simply asserts that those who refrain from lying secure a good reputation. Because it asserts a causal link between doing the means and actualizing the end, Kant says that it concerns the grounds ‘for actualizing the object’. This must be distinguished, however, from the practical judgement. This asserts a connection between the very same objects but also connects them to my willing. Instead of saying that those who tell the truth tend to have a good reputation, it says that if securing a good reputation is the thing to do then telling the truth is also the thing to do.Footnote 15 As Kant puts it, it asserts a connection between grounds for actualizing the will.
In Lying-H, it is only the practical connection, the one that concerns the ground of willing, that is analytic. We might put Kant’s point this way. Given the synthetic theoretical connection that exists between lying and ruined reputations, there is an analytic practical connection between willing to keep my reputation and willing honesty. The theoretical connection between lying and ruined reputations is a synthetic, empirical one. Given this, however, the practical connection between willing a good reputation and not lying is one that relies only on rules constitutive of practical thinking. In other words, an agent who accepts the theoretical connection between lying and ruined reputations, and wills a good reputation but refuses to will the means to this, is not just reasoning badly, but is, in a sense, not practically reasoning at all.
But the situation is different with Lying-C. Although Lying-C will also involve empirical synthetic propositions (for example, the meaning of English phrases and facts about what the person with whom I am communicating is likely to believe), in Lying-C telling the truth is not an analytic consequence of anything else. Calling an imperative p-synthetic just means that even after we throw in all the empirical synthetic claims Lying-C, unlike Lying-H, does not follow by means of rules constitutive of practical reasoning.
The contrast can be made clear if we do not think in terms of hypothetical and categorical imperatives but instead in terms of hypothetical and categorical reasoning. Consider the following instance of hypothetical reasoning:
-
1) I keep my reputation. (Willed premise)
-
2) Lying ruins my reputation. (Empirical, synthetic premise) _______________
-
3) I do not lie. (Willed conclusion)
versus:
-
4) I keep my reputation or I do not keep my reputation. (Willed premises) ________________
-
5) I do not lie. (Willed conclusion).
Put this way, Kant’s belief that 1–3 is analytic amounts, according to P-Analytic*, to the claim that 3 is entailed by 1 and 2 by means of rules constitutive of practical agency as such. Kant’s belief that 4–5 is synthetic, on the other hand, amounts to the claim that 5 is not entailed by 4 by means of rules constitutive of practical agency itself. Since 4 is a tautology, Lying-C states that no matter what we will, 5 follows as a matter of necessity. That 5 follows from no premises makes it look like a tautology, which perhaps makes it sound like a logical law. Appearances here are deceiving. Although 5 is entailed by anything at all, it is central to Kant’s thought that it does not follow by means of rules constitutive of practical reasoning. It is entailed but not constitutively entailed. The important difference between 1–3 and 4–5, therefore, is not the number of premises but the kind of rules according to which the conclusions follow from these premises. According to Kant, 3 follows from rules constitutive of practical reasoning whereas 5 follows from some other type of rule. The two sets of practical reasoning, therefore, differ with regard to the inference rules not the number or kinds of premises.
Modal Dualism
The contradiction characterization has the virtue of bringing out into the open what Hanna calls Kant’s ‘modal dualism’ (Hanna Reference Hanna2001: 26–7) – the belief that there are two different kinds of necessity. What I am suggesting here is that Kant’s application of ASD to the practical sphere commits him to a kind of practical modal dualism. Some necessary practical judgements are vindicated by rules constitutive of practical reasoning as such and some necessary practical judgements are not.
It is important to distinguish Kant’s claim that there are two different kinds of practical necessity from his claim that there are different modes of practical necessity.Footnote 16 When Kant introduces the distinction between hypothetical and categorical imperatives, he does so by noting that hypothetical imperatives are necessary but only on the condition of a pre-existing desire, whereas categorical imperatives, depending on no desires, are unconditionally necessary. But this difference between conditional and unconditional practical necessitation is not a difference in kind. Consider the principles that govern ordinary deductive logic. Even though it only recognizes one kind of necessity, it still distinguishes between beliefs that are conditionally necessary (you must believe p provided you believe q) from the unconditionally necessary beliefs (you must believe ‘p or not p’). Whether the rules apply conditionally or unconditionally is a different question from how many kinds of rules there are. When Kant says that hypothetical imperatives are analytic and categorical imperatives are synthetic, he is going beyond the distinction between conditionally and unconditionally necessary. He claims not only that a rule of practical inference may be differently applied but that there are (at least) two separate rules of practical inference.
The distinction between modes and kinds of necessity sheds light on an important if overlooked puzzle about hypothetical and categorical imperatives, i.e. that some hypothetical imperatives seem to be able to be transformed into categorical imperatives. Suppose, for example, that we believe that lying ruins our reputation. Surely, it follows that we ought disjunctively to: either forgo lying or forgo our reputation. But this seems to be required of us independently of anything that we happen to desire. If we understand categorical imperatives to be those that apply to us unconditionally, then this would be a categorical or moral imperative. When Kant applies ASD to imperatives, he gives us a way out of these sorts of dilemmas. The difference between hypothetical and categorical imperatives is a strong difference in kind. P is analytic because we can see that it is required using only rules that are constitutive of rationality as such, and it remains so even under logical translations.
There is a price for thinking that there are two kinds of necessity only one of which is grounded in the rules constitutive of practical thinking. Since morality clearly belongs to the non-constitutive kind of necessity, it follows that the necessity of morality cannot be understood by appealing to what is constitutive of our practical thought. But this is just to say that Kant is not an austere constructivist.
5. Kantian Constructivism Considered
I have argued that there are textual reasons to doubt that Kant was an austere constructivist. At the outset of this article, however, I also argued that Kant’s denial of austere constructivism does more than just correct the historical record – it also opens up an exciting possibility within constructivism itself. Although it is impossible, here, to provide a full articulation and defence of authentic Kantian constructivism, I would like to briefly sketch what such a position might look like and show that it fits in neither with Humean nor austere constructivism.
Kantian constructivism starts with the claim that the moral law is a synthetic a priori practical judgement. In the theoretical realm, synthetic a priori judgements are justified when they are shown to be necessary for representing an object of experience. Allison calls the principles which allow us to vindicate these synthetic a priori judgements ‘objectivating conditions’ (Allison Reference Allison2004: 11) since they are the principles which make possible objective representation (i.e. representation that is about an object). If analytic truths derive from formal conditions of thinking, synthetic truths derive from formal conditions of experience (A156–7/B196). For example, to vindicate his claim that ‘all alterations occur in accordance with the law of the connection of cause and effect’ (B 232), Kant tries to show that the causal principle is necessary for representing temporally located objects and then argues that temporally located objects are part of any objective experience (i.e. experience that is about an object). Whether or not one thinks that Kant is successful, this method of arguing is clearly different than checking to see whether the judgement is a logical tautology.
As I interpret it, Kant’s decision to apply ASD to the practical sphere suggests an analogous strategy for moral judgements. Here too Kant would start by distinguishing between the conditions of practically thinking and the conditions of practically representing an object. The former would be formal, practical and pure philosophy,Footnote 17 but it would have nothing to do with the argument in the Groundwork which is supposed to be a groundwork of material practical pure philosophy, i.e. a metaphysics of morals. In the Groundwork, we should expect to find an argument that the moral law is an objectivating practical condition, i.e. a condition necessary for representing a practical object. There will be no austere arguments purporting to show that the Categorical Imperative follows from the bare idea of practicality as such. Instead, the Categorical Imperative will be vindicated because it is necessary for representing objective practical objects.
Such an argument requires an understanding of the idea of a practical object and what it is to represent such an object. In the Critique of Pure Reason, Kant spends a lot of time expounding the difference between a subjective and an objective representation (e.g. B142) so that by the time he sets about arguing for the pure principles of understanding, we know what it would mean to show that a principle was necessary to achieve an objective representation. But there is no corresponding analysis in the practical works. In an insufficiently discussed place in the Critique of Practical Reason, Kant does say that the good and the evil are ‘the only objects of a practical reason’ (CPrR, 5: 58). This indicates that the difference between a subjective practical representation and an objective one is that the latter is represented as good. But Kant does not explain how this understanding of a practical object relates to his understanding of a theoretical object such that we could understand them as different species of a common genus. Nor does he explain here how we are to understand representation in the practical sphere. Now I do not mean to say Kant does not answer these questions elsewhere. What is important is just to admit upfront that the problem of the synthetic practical a priori is not treated as explicitly or centrally as it is in the theoretical philosophy.
Still, I think that enough has been said to show how different a constructivism that was built around the synthetic practical a priori would be from either austere or Humean constructivism. To see the difference let us consider again Street’s ‘ideally coherent Caligula’ (Street Reference Street2010: 371). Recall that according to Street the austere constructivist must think that, contrary to appearances, Caligula really does have reason not to value torturing others for fun and these reasons derive from facts about agency as such. But Kantian constructivists would insist that formal features of our practical agency are as empty as Kant thinks theoretical logic is. Instead, they would want to know what sort of practical agent Caligula was. In particular, they would want to know whether or not Caligula has any objective practical representations. Was he capable not only of desiring things and looking to find the means to these desired objects or did he hold that the things he desired were in fact objectively good? If Caligula is only a minimally rational agent, in the sense that he has desires but does not hold that any of them are good or important, then he will have no reason to abandon torturing for fun. If, on the other hand, he holds that torturing is objectively good, then a Kantian constructivist will say that, contrary to appearances, he does have reason not to value torturing. The austere constructivist tries to derive substantive conclusions from the formal conditions of agency as such. The Kantian constructivist attempts to show that if Caligula succeeds in representing the targets of his desires as objectively good, then he will be subject to the moral law.
But the Kantian constructivist is not a Humean constructivist either. While the Humeans are right to note that there will be some agents such that no arguments could bring them into morality, they are wrong to think that disagreeing with the torture of others is just a contingent matter of happening to value morality. Morality is a necessary condition of representing some things as objectively practical. Any creature who can distinguish between the subjectively pursued and the objectively good must be subject to morality. Moral judgments are not derived from some bare idea of rationality as such, but are instead derived from a robust understanding of a practical agent capable of representing things as objectively good.
6 Conclusion
An austere constructivist believes that the moral law is a constitutive principle of rational agency as such. Such an interpretation explicitly conflicts with the preface to the Groundwork. But perhaps more importantly, such an interpretation conflicts with Kant’s decision to apply ASD to the practical sphere. In the theoretical sphere, ASD distinguishes between two types of necessity, necessity that is constitutive of thinking as such and necessity that goes beyond such constitutive considerations. If ASD in the practical sphere is also going to distinguish the merely constitutive from what goes beyond the constitutive, then Kantian morality clearly goes beyond the merely constitutive.
In the Prolegomena, Kant calls on people who work on metaphysics to stop whatever it is they are working on until they have answered the question ‘how is metaphysics possible?’ (P, 4: 255). What is more, he insists that answering this question involves understanding how synthetic a priori judgments are possible (P, 4: 276). In the light of this high rhetoric, Kant’s application of ASD to imperatives is quite striking. Perhaps the possibility of a metaphysics of morals also depends upon understanding how practical synthetic a priori judgments are possible. Perhaps the central question of practical philosophy involves distinguishing practical analytic from practical synthetic necessity. If this is right, then following the lead of many Kant commentators in interpreting Kant as an austere constructivist means that we will never even start addressing what Kant would have thought is the central problem of practical philosophy.Footnote 18
