Kant is known as a champion of the idea that moral conduct is demanded by reason; but, despite a remarkable revival of interest in Kant’s ethics, surprisingly little attention has been paid to Kant’s explicit argument for the idea.1 This neglect is understandable but unfortunate. The misfortune is not that we have overlooked a sound and lucid proof which could have effectively settled all contemporary controversies about whether it is rational to be moral; it is rather that we have missed, or misread, a text which is crucially important for understanding Kant’s Groundwork of the Metaphysics of Morals and which contains ideas worth considering in their own right. The argument in question is what is summarized in the opening paragraphs of the notorious third chapter of the Groundwork. The usual reading not only makes Kant appear careless and unbelievably confused; it reinforces an interpretation which has Kant holding the outrageous view that immoral acts are unfree and not even willed. Moreover, common readings of the argument have Kant arguing from a forbidden empirical premise (regarding a feeling of freedom), confusing natural laws and laws of conduct, and committing an obvious non sequitur by overlooking the fact that he can only prove freedom “from a practical point of view.”
In what follows I will sketch a reconstruction of the main features of Kant’s argument, a reconstruction which I believe avoids these gross difficulties and yet remains (largely) faithful to the text. My discussion, however, will concentrate more heavily on the earlier stages of the argument, for two reasons. First, those early stages, in which Kant argues that rational wills have autonomy, offer an intriguing proof that Humeans and Hobbesians are mistaken about the nature of practical reason, and the proof is quite independent of Kant’s belief that he has identified the supreme moral principle. Second, though the earlier stages have drawn the heaviest ridicule, I think they are both more crucial to the interpretation of the Groundwork and more promising than the later stages. Though focusing upon the earlier parts of Kant’s argument, I will however comment briefly on a step at the end where I believe Kant goes wrong.
The task Kant undertakes in chapter 3 of the Groundwork is nothing less than proving that moral constraints are requirements of reason. His argument, then, amounts to an answer to the contemporary question “Why be moral?” But Kant’s aim is easily obscured by the fact that his imagined audience is not the sort of moral skeptic with which we are most familiar today. Kant does not see himself as addressing, for example, those who are indifferent to morality and demand that philosophy supply them with a motive to be moral; for Kant’s own theory denies that anyone rational enough to ask the question could really be so indifferent.2 Nor, I think, is Kant addressing an audience that doubts that common sense duties, as opposed to some revisionary standards, are genuinely moral (G, 403-4). He does not imagine that anyone who clearly understood his supreme moral principle would need to wait for a proof before he felt its rational force. The intended audience, I think, is rather those whose moral commitment is liable to be called into question by philosophical accounts of practical reason which imply that morality could not be grounded in reason. To these Kant argues, first, that their theories of practical reason must be mistaken and, second, that the only alternative shows moral requirements to be rational. The argument, if sound, has important implications for contemporary moral skeptics; but its focus, its style, and perhaps even the degree of care devoted to its parts are influenced by Kant’s own conception of his audience. That audience may have picked up ideas of Epicurus, Hobbes, and Hume; Nietzsche comes later.
The aim of the third chapter, in Kant’s terms, is to establish the supreme moral principle. This is the second of two main aims stated in the preface of the Groundwork, namely, “to seek out and establish the supreme principle of morality” (G, 391-92). Seeking out, or identifying the principle, is accomplished in the first two chapters, which also analyze the principle, reformulate it in various ways, and relate it to ideas of moral worth, dignity, etc. In the second chapter, using a so-called “analytical” method, Kant also argues that the concept of moral duty presupposes unconditional commands of reason not based on desires and hypothetical imperatives; but this only raises the stakes for the third chapter and leaves us with the possibility that there may be no genuine moral duties. That is, the argument purports to show that if there are moral duties then there must be non-desire-based requirements of practical reason (G, 425); but whether or not there are such rational requirements is left an open question. As Kant says, for all that has been shown, morality may be “a phantom of the brain” (G, 445), that is, a set of constraints falsely believed to be rational but actually having their source in imagination rather than reason. In chapter 2 the various formulations of the supreme moral principle are labeled “the Categorical Imperative” (G, 421, 428-29, 440), but it is admitted that no proof has yet been given that they are categorical imperatives or indeed that a categorical imperative is possible (G, 425, 440, 445). That task is left for chapter 3.
If my reading is right, the argument of the third chapter is obviously important; why, then, has it been so often overlooked or maligned? The most obvious explanation lies in the fact that the argument is extremely compact, unclearly stated, and deeply entangled with aspects of Kant’s metaphysics that have little appeal today. A further obstacle is that Kant himself suggests that he may have been reasoning in a circle (G, 440-41), and though he claims to have found a way out of the circle this turns on an introduction of the “intelligible world” that is not obviously helpful (G, 453). One is discouraged from trying to unravel all this by the fact that in the Critique of Practical Reason Kant seems to abandon the project of establishing the rationality of morals. There moral obligation is simply declared a “fact of reason” and used to establish that rational wills are free (which is a crucial premise in the argument of the Groundwork).3 Another obstacle is the fact that Kant spends so much of the third chapter of the Groundwork stressing the compatibility of phenomenal determinism and noumenal freedom that one is tempted to see the point of the chapter as a defense of morality against determinism. Finally, I suspect that Kant’s argument has been underrated because many sympathetic commentators believe that, on Kant’s own principles, it is unnecessary and perhaps even morally corrupt to ask seriously, “Why be moral?”4 To read the third chapter of the Groundwork as an attempt to answer this question, then, would be to see it as misguided and bound to fail.
Although these considerations help to make the neglect of Kant’s argument understandable, they are not, I think adequate to justify it. One can make some headway despite the obscurities and heavy metaphysics; the reversal of premises and conclusion in the Critique of Practical Reason can be explained by the different nature of its project; 5 and, despite Kant’s dramatic rhetoric about circular reasoning, the argument of the third chapter is not in fact a circular one.6 The long discussion of the compatibility of freedom and determinism (in chapter 3) cannot be the main point, because this was supposedly demonstrated earlier in the Critique of Pure Reason and it would not answer the question so provocatively declared open at the end of chapter 2, namely, whether alleged duties are, as they purport to be, genuine unconditional commands of reason. Again, while Kant thought it a mistake to try to give reasons for being moral in terms of desired ends contingently served by morality, this does not mean that he failed to recognize a need to demonstrate that moral requirements are rooted in reason.7
The most crucial passages are the following:
Will is a kind of casuality belonging to living beings so far as they are rational. Freedom would then be the property this causality has of being able to work independently of determination by alien causes. . . .
The above definition of freedom is negative and consequently unfruitful as a way of grasping its essence; but there springs from it a more positive concept, which, as positive, is richer and more fruitful. The concept of causality carries with it that of laws (Gesetze) in accordance with which, because of something we call a cause, something else—namely, its effect—must be posited (gesetz). Hence freedom of the will, although it is not the property of conforming to laws of nature, is not for this reason lawless: it must rather be a causality conforming to immutable laws, though of a special kind; for otherwise a free will would be self-contradictory. . . . What else then can freedom of the will be but autonomy—that is, the property which the will has of being a law to itself? The proposition ‘Will is in all its actions a law to itself’ expresses, however, only the principle of acting on no maxim other than one which can have for its object itself as at the same time a universal law. This is precisely the formula of the Categorical Imperative and the principle of morality. Thus a free will and a will under moral laws are one and the same. (G, 446-47).
This passage argues in effect that, if free in a negative sense, every rational will is committed to morality. The following passage contains the nub of the rest of the argument, which is to show that every rational will is free (in a negative sense).
Now I assert that every being who cannot act except under the Idea of freedom is by this alone—from a practical point of view—really free; that is to say, for him all the laws inseparably bound up with freedom are valid just as much as if his will could be pronounced free in itself on grounds valid for theoretical philosophy. And I maintain that to every rational being possessed of a will we must also lend the Idea of freedom as the only one under which he can act.8
The main outline of the argument is clear enough. Reordering the parts, we have (1) an argument that any rational will is free in at least a negative sense, (2) an argument, turning on definitions of negative freedom and autonomy, that any will free in the negative sense has the property of autonomy (G, 446), (3) an assertion, relying on earlier arguments, that any rational will with autonomy is committed to the principle “Act only on maxims you can will as universal laws,”9 (4) an assertion, again relying on earlier arguments, that this last principle is the supreme principle of morality.10 From all this it follows that any rational will is committed to the supreme principle of morality. Thus we can conclude that anyone who acts immorally is acting contrary to a principle which he himself accepts. If we assume that the principle in question is an unconditional and not merely prima facie one,11 then such immoral acts must be irrational because contrary to an unconditional commitment of the agent. If we assume further that the argument shows that rational autonomous agents are committed to the supreme moral principle qua rational and autonomous, then immoral acts will be irrational not only because they are contrary to the agent’s commitments but also because they are contrary to a principle the agent acknowledges to be rational.12
Each of the major steps, of course, depends upon subsidiary arguments, some of which I shall consider shortly. But before that, a few preliminary remarks may be helpful.
First, in steps (1) and (2) Kant aims to show that any rational will has autonomy, and these stages presuppose nothing about morality. Now though these stages are parts of a larger argument that it is rational to be moral, they are of interest independently of that larger project. They are concerned not merely to affirm a freedom unthreatened by causal determinism but to argue the necessity of acknowledging rational principles of conduct other than those which prescribe efficiency in satisfying our desires or in coordinating our means and ends. The conclusion, even at this earlier stage, is that anyone who acts for reasons must acknowledge at least some reasons other than facts about what is needed to achieve his ends and to satisfy his desires. If so, then regardless of what we think of morality there must be practical reasons which are not hypothetical imperatives. Hobbesians will be wrong to construe all rational principles as rules of rational self-interest; modem decision theorists will be wrong if they suppose that all rational choice principles are relative to intrinsic preferences themselves uncriticizable by reason; and Humeans will be wrong to suppose that reason merely calculates and discovers facts rather than prescribing conduct. The conclusion is a strong and controversial one: a striking feature of the Groundwork is that Kant’s argument for it does not depend upon claims about morality.
As a final preliminary, I must say a few words about the interpretation which I shall try to avoid.13 According to this, the will is practical reason and so cannot will anything contrary to reason; morality is prescribed by reason and so no one wills to be immoral; the will, which is thus always good, is free negatively and wills unequivocally perfect conformity to the laws of autonomy. Thus, on this view, one who acts to satisfy desire contrary to morality, and perhaps even one who acts to satisfy a morally neutral desire, does not really will so to act and does not act freely in any sense. His behavior is a product of natural forces, like that of animals or, better, animals with complex built-in computers for calculating the best means to satisfaction. We are strange hybrids sometimes governed by freely acknowledged rational moral principles and sometimes in the grip of natural forces beyond rational control; and what switches us from the one mode to the other is inexplicable. It could not be a free choice because to be capable of free choice is to be in one mode rather than the other. When we act from desires we act heteronomously, which is to say unfreely and nonrationally; when we act from moral principle, we act autonomously, which is to say freely and rationally. And there can be no free choice between the two, for free choice is always for the rational and moral.
Kant does say things in the Groundwork which suggest this strange picture, but I am convinced that the textual evidence, on the whole, is opposed to it. However, in order to lay out my reconstruction of the argument under consideration even in the present sketchy form, I must leave detailed examination of particular passages for another occasion. For now I must be content to offer an alternative reading which, I hope, makes more sense of the compressed and puzzling argument in the third chapter and to call attention to the disparity between the interpretation I reject and the views Kant makes more evident in his later ethical writings. These later works,14 with the explicit distinction between Wille and Willkür, make clear that Kant then thought the adoption of ends in general and certainly the adoption of immoral maxims were free choices of a rational agent, even though not maximally rational choices. To understand the Groundwork in the context of Kant’s work as a whole without regarding it as a radical deviation, we must see if we can understand his argument without having to attribute to him the bizarre picture I have sketched with its consequence that immorality is unfree and unwilling.
Conceptual analysis, Kant suggests, should suffice to show that any rational will which is free in a negative sense is also free in the positive sense (autonomy). The crucial definitions are those of will, negative freedom, and autonomy.
Will is a “kind of causality,” distinct from causation by prior events and natural laws, a sort of ability to make things happen peculiar to rational agents (G, 446). Elsewhere, importantly, Kant characterizes the will as a power to act “in accordance with (his) idea of laws—that is, in accordance with principles” (G, 412, 427). The idea is that to be an agent, or a rational being with a will, one must be able to make things happen in such a way that the appropriate explanation is reference to the principles, laws, or reasons on which the person acted. Principles, even laws, enter into the explanation of why a rational agent did something (as distinct from merely why the body moved) as the agent’s guiding “ideas” or rationale, not as empirically observable regularities among types of events. In fact the will for Kant (in contrast, say, to Hobbes and Hume) is not an event, a mental episode occurring prior to action, which explains that action in the ordinary empirical way.15 Kant believed, of course, that explanations of an act by reference to the agent’s reasons (and so his will) were compatible with accepting deterministic explanations of the corresponding behavior by empirical laws; but that belief is not essential to the important distinction between the two types of explanation.
This conception of the will has several important implications. First, Hume’s famous reductio ad absurdum of indeterminism does not apply to Kant.16 Hume, assuming that an indeterminist’s “free will” was an uncaused event prior to an agent’s act, argued that such an event, unconnected with the agent’s character, was not something for which a person could be held morally responsible. Kant undercuts the objection by denying that “willing” is a prior event. His own metaphysical account of the will has problems enough of its own, but he may be right to suspect that, in its classic form, the dispute between determinists and indeterminists rests on a shared model of rational action that is inadequate. As many since Kant have acknowledged, those who are troubled by the picture of desire, act of will, and bodily movement as discrete physical events in a causal sequence, like falling dominos, do not obviously render responsibility less puzzling simply by denying the causal connection between the first two items (desire and will) and/or by assigning these items to an introspectible mental realm. Second, since an act of will for Kant is not an introspectible phenomenon, it is no reply to Kant’s argument that the will is committed to a certain principle (e.g., the Categorical Imperative) to say, “But I don’t remember deciding to follow that.” An argument that the will of every rational agent is committed to morality need not be based on observations of their life histories. In at least this respect moral commitment is supposed to be like rational commitment to basic principles of logic and empirical understanding. Third, behavior cannot be attributed to the will of an agent, not even to the “free will” of an agent, unless it is supposed that the agent was acting for a reason, or guided by “the idea of a principle (or law).” Thus it is part of the concept of a will that it cannot be “lawless.”17
Next we need to look at the negative concept of freedom, which is “the property (the will) has of being able to work independently of determination by alien causes.” It is clear enough that Kant means at least to deny that there is an empirical causal account for why free wills act for the reasons they do. To attribute an act to the free will of a rational agent is not to cite its empirical causes or to refer to an empirical mechanism or power, caused or uncaused, which explains how an observed behavior came about. The Critique of Pure Reason is supposed to have established the possibility that such a will is in some sense responsible for what we do despite Kant’s insistence, and supposed proof, that empirical science can in principle give causal explanations of all phenomena, mental as well as physical. The first Critique also makes clear the price Kant is willing to pay for this compatibility, namely, conceiving the will as something apart from the spatiotemporal order which we can “comprehend”; but that full price, I think, is not essential to the main thrust of our argument.
The preceding remarks reflect the usual (though still vague) understanding of a will’s capacity “to work independently of determination by alien causes.” But, importantly, there must be more than this to Kant’s conception of negative freedom if the step from negative freedom to autonomy is at all plausible. For even if we conceive a negatively free will as somehow independent of causal determinism, we could still regard such wills as (causally inexplicable) capacities to act for the sake of satisfying desires and inclinations. In other words, though acting for reasons is not to be understood as being causally determined by one’s given desires, nevertheless one’s capacity to act for reasons might be limited, perhaps even by the concept of reasons, to policies aimed at satisfaction of some of the desires and inclinations one happens to have. We might speak of such a will as incapable of motivation by anything but inclination and desire, where “motivation” refers not to what causes the willing but rather to the range of things the will can count as reasons or rational objectives of its policies. Though again I must forgo detailed textual argument, I think it is evident once the distinction is made that Kant regarded a negatively free will as also capable of acting independently of motivation by desires and inclinations. There is a sense, perhaps unfortunate, in which Kant regarded even the agent’s own desires as “alien,” and “determination” of the will, even by “alien” factors, does not refer exclusively to having a place in a deterministic nexus of causes. On the contrary, when Kant writes of a will “determined” by reason, this is not to cite a prior event and a causal law but rather to say that the guiding idea on which the agent acted was a rational one; and, similarly, when a will is “determined” by inclination in a standard case (not a knee jerk, reflexive scratching, etc.), this means the agent’s policy or guiding idea was some hypothetical imperative concerning the means to satisfy the inclination. In the latter case there is a (misleading) sense in which “alien causes” determine the will; it is not that the agent’s inclination deterministically causes the agent to will what he does, but rather that the agent’s chosen policy makes a certain causal connection, or strictly his belief in a certain causal connection, be a decisive (or “determining”) factor for what he does. His full rationale (not a causal event) is: “I shall do whatever is necessary as a means to satisfy my inclination B; A is a necessary means to satisfy B; hence I shall do A.” The agent has let the causal law, or strictly his idea of the causal law, between that sort of means and end be the dominant or “determining” factor in his choice; but this does not mean that the willing itself was subject to causal explanation.
Notice that the interpretation I have pledged to avoid does not make this distinction between being caused to act by one’s inclination and choosing to act on policies which make satisfaction of inclination the rationale for acting. Once it is made, however, it seems clear that Kant’s conception of freedom, even negatively defined, encompasses not only capacity to will without the willing being explainable by causal laws and prior events; crucially, freedom also includes the ability to will, or act for reasons, where the agent’s rationale is not a hypothetical imperative indicating the means to satisfy an inclination. Without this stipulation the argument that negatively free wills necessarily have autonomy would fall flat: for autonomy, as we shall see, implies a capacity/disposition to follow principles other than desire-based hypothetical imperatives.
The most difficult definition in the argument is that of autonomy (or freedom positively conceived). Part of the difficulty in understanding this stems from the fact that the misguided picture I am trying to avoid is rather deeply entrenched in commentaries and is encouraged both by ambiguities in Kant and by everyday (non-Kantian) talk about autonomy. On that picture autonomy is an ideal achieved by some and not others, or perhaps by some people some of the time but not always. Rather than a property of all human wills, it is seen as a property of purely conscientious wills when willing out of respect for the moral law. The misleading picture is reinforced by a facile use of the expressions, which are not Kant’s, of “acting autonomously” for free/morally inspired conduct and “acting heteronomously” for causally determined or at least desire-motivated conduct.
To the contrary, Kant’s view, I think, was that autonomy is a property of the will of every minimally rational agent, which includes virtually all adult sane human beings, no matter how wicked. Heteronomy is a possible property of wills which misguided moral philosophers have mistakenly attributed to human beings.18 All rational agents, Kant argues, have negative freedom, and to have negative freedom is to have autonomy as well. This is not to say that everyone chooses to fulfill the commitments he has by virtue of having autonomy of will. The immoralist is not one who has a will characterized by heteronomy but rather one who acts as if the human will were such, i.e., one who in practice ignores the implications of having a will with autonomy and acts as if the only authoritative rational principles were hypothetical imperatives.
Kant’s most explicit definition of autonomy is that it is a will’s property of “being a law to itself (independently of every property belonging to the objects of volition)” (G, 440). Though by itself this is not so illuminating, what is meant, I think, can be plausibly reconstructed as follows: a will with autonomy is not only negatively free but is committed to at least one principle acknowledged as rational to follow but such that (a) one is not causally determined to accept or follow it, (b) it does not merely prescribe taking the necessary (or best) means to one’s desired ends, (c) the rationality of accepting it does not depend upon contingent facts about what means will serve one’s ends or about what ends one happens to desire, and (d) the principle is “one’s own” or “given to oneself by oneself,” i.e., it expresses a deep commitment from one’s “true” nature as a rational and (negatively) free agent rather than, say, expressing respect for an external authority, tradition, conventions, etc. All this is simply a long way of saying what is (more vaguely) summarized by saying that a will with autonomy accepts for itself rational constraints independently of any desires and other “alien” influences.
With these preliminaries and definition in hand, we can see how that argument from negative freedom to autonomy must go. Negatively free rational wills can act for reasons without being motivated by desires and hypothetical imperatives. But, as they are not “lawless,” when they so act they must be following some principle (or principles) which allows us to attribute the act to a rational agent, acting for reasons, as opposed to whimsical behavior, knee-jerk reactions, etc. So the agent must have, or be committed to, principles he acknowledges as rational even though they are not of the hypothetical imperative sort. Because the agent is negatively free, acceptance of such principles cannot be causally determined. Since principles adopted because of external authorities (e.g., God), tradition, convention, etc., would be based on hypothetical imperatives, these cannot be the rational principles in question. The only alternative, it seems, is that the principles reflect some necessary features of rational agency itself independently of its special contexts. If we assume (with Kant) that one’s nature as a rational will is in some sense one’s “true” self in contrast to passively “given” phenomenal desires, then we could conclude further that the rational principles in question are “one’s own” or “given to oneself by oneself” in a way that desire-based principles are not.
The last step raises deep questions about personal identity which need not cloud the main point, namely, that if one can act rationally without causal determination and without following desire-based principles, then one must have some principle or principles which are rational and yet not hypothetical imperatives. If rational agents are negatively free, then, they must acknowledge rational principles of conduct beyond those recognized by followers of Hume and Hobbes. This conclusion is reached not by exhibiting an example of such a nonhypo-thetical rational principle but by indirect argument that there must be such if there is rational free agency.
Since the argument turns on the capacity to act without motivation by desire, the conclusion is not that free agents invariably follow nonhy-pothetical rational principles but only that they must have or acknowledge them. If they had no such commitments, they would lack the ability to act rationally without following hypothetical imperatives; but having the capacity does not mean that they will follow the principles whenever they can (or even whenever the principles prescribe for them). Thus, when later stages of the argument identify the laws of autonomy with moral constraints, it will not follow that free rational agents invariably act morally but only that they can (and must to avoid irrational conflict of will).
It may be useful to consider two objections. First, does it make sense to say (with Kant) that rational agents can act independently of desires and inclinations? This depends upon how these are conceived. Kant’s idea of desires and inclinations, at least in the context of our argument, is narrower than some conceptions and wider than others. If it is confused with the narrower conceptions, Kant’s idea of freedom will seem less controversial (and interesting) than it is; and if Kant’s idea of desires and inclinations is confused with the wider conceptions, then his idea of freedom will seem absurd. To begin with the wider conception, there seems to be a use of “desire” and related words which refers to whatever motivates an agent, whether empirically discernable independently of his act or not. That an agent “desired” (or, better, “wanted”) to do something in this sense is simply inferred from the fact that he did it, given options. Since there are no independent grounds for attributing the “desire,” obviously mentioning it does not explain what was done but only characterizes it as voluntary, unlike knee jerks, etc. To say that we can freely will to act independently of desire, in this sense, is obviously absurd; and Kant did not mean this, for he held that even the purest moral acts are motivated, in some sense, by an inexplicable “interest” in the commands of reason.
Sometimes “desire” (and related words) seem to be used quite narrowly, referring to dispositions that are noticeably (and often urgently) felt by the agent prior to acting, involving pleasant anticipation, painful fear of loss, tendencies to search for means of fulfillment, to experience frustration when thwarted and joy when fulfilled, etc. It is tempting to suppose that this is what Kant had in mind, for it is not very controversial that we can act for reasons without being motivated by desires in this narrow sense.19 We seem to do so often when we go about our routine business, forgo minor pleasures for health reasons, and so on. Kant does at times seem to have in mind this narrow sense, especially when dramatically depicting struggles between duty and inclination; but the argument needs more than this and Kant clearly intended more than this in his discussions of freedom. We would still be acting independently of desire in this narrow sense whenever our reasons were based simply on Hume’s “calm passions,” or mild preferences unaccompanied by pleasures of anticipation, or aversions we attribute to ourselves more from inference than from feeling. But, however, rational, such acts would not have the independence of empirical motivation Kant attributes to acts which manifest freedom.
What, then, is the relevant sense of “desire” and “inclination”? Fine points aside, these are virtually any empirically discernible motivations that one may happen to have so long as they are not concerns essential to all rational agents as such. They include desires in the narrow sense but also Hume’s “calm passions” and any other preference, liking, aversion, love, hate, etc. which rational agents might lack and which is not attributed solely because they acted voluntarily. Negative freedom and autonomy imply capacity for rational action in which the agent’s reason is not to achieve or do something desired in this intermediate sense. Since Kant repeatedly grants that any behavior may also be empirically explained (e.g., by desires, by the feelings associated with “respect,” or by physical science), his point is not to deny this but to say something about reasons: namely, that the status of the rational agent’s end as a reason for the agent does not always depend upon the agent having towards it an empirical disposition of the sort rational agents might or might not have. The thesis that we are free agents in this sense is controversial, of course, but not obviously absurd (as on the wide conception of desire) or uninterestingly true (as on the narrow conception of desire).
A second objection should be mentioned.20 For all that has been said, is it not possible that a will with autonomy has for its one and only rational principle the pure Hypothetical Imperative? This is the principle that it is rational to take the necessary and available means to one’s ends or else abandon those ends. I call this “the Hypothetical Imperative” because it is a version of the principle behind Kant’s particular hypothetical imperatives, and I call it “pure” because it does not specify that the ends are willed because desired, or for some other reason, or indeed for any reason. It might seem that my characterization of autonomy leaves open that this is the only non-desire-based rational principle.
One problem with this suggestion is that the principle in question (as the name suggests) does not unequivocally prescribe any act even given full information about the situation, one’s preferences, etc. It says only, “Take the means or drop the end,” giving no standards for the rational assessment of ends. Now conceivably this principle (in a more carefully stated version) is the only necessary principle of rational agency; but then agents who followed it, selecting their ends according to their inclinations or for no reason at all, would never act for sufficient reasons, as (I believe) Kant thought a rational free will could do. Admittedly, in a sense such a will would not be “lawless,” for it would have the pure Hypothetical Imperative as a necessary rational principle; but, unless we assume something further, reason would at best prescribe an option rather than a course of action.21 Though Kant does not raise the issue explicitly, I take it that he conceived rational free wills as sometimes more definitely constrained than this.
If we could assume (as many do) that all preferences, or all preferences that survive a process of informed reflection, have some weight as reasons, then ends could be rationally assessed by the likelihood and costs of achieving coherent sets of preferences. But this would presuppose a putative principle of rationality beyond the pure Hypothetical Imperative, which is the only principle of practical reason Kant recognizes as analytic. Since that further principle assigning prima facie rational force to our empirically given preferences (or informed preferences) is nonanalytic, unless it could be given a “synthetic a priori” justification Kant would deny it status as a necessary principle of reason just as he denies this status to the substantive principles “Satisfy the promptings of your (supposed) moral sense” and “Do all you can to satisfy your desire to be happy.” Assuming, then, that such principles are not necessarily rational and also that free wills independently of their given desires and inclinations can nonetheless have sufficient determinate reasons to act, one whose “reasons” were fixed solely by the pure Hypothetical Imperative and de facto preferences (or informed preferences) would not have a will with autonomy.
These remarks reveal a heavy burden left for the next stage of the argument, i.e., the attempt to show that rational wills necessarily have negative freedom (and so autonomy), but we should perhaps not expect the conditions for freedom to be so easily satisfied that the concept is of little use in Kant’s ultimate project of showing that the basic standards of morality are necessarily rational.
Now, more briefly, we need to consider the notoriously compact argument that every rational will is (negatively) free.
First some preliminary comments. Though Kant does not explicitly refer to freedom negatively conceived in presenting the argument, this is what the argument requires (given that the transition from negative freedom to autonomy is argued separately). Further, we should not think of the argument as concerned with some new sense of freedom (“practical freedom”), as some have suggested; the qualification “from a practical point of view” which Kant attaches to the argument refers not to a sense of freedom but to the type of argument given and a restriction on the legitimate use of its conclusion. Again, we should note that the argument, strictly speaking, is not that we, or any particular individuals, are free; it is that every rational agent, as such, must be free. Later, after raising the objection that his reasoning may have been circular, Kant supplements the basic argument under consideration with the contention that even in our theoretical judgments, apart from practical questions, we must take ourselves to be members of an “intelligible world,” in some sense independent of given sensuous inclinations (G, 448-53). This further argument seems relevant to residual doubts that we are rational free agents (as defined); but this goes beyond the project at hand. Finally, the standards for rational agency must not be too high. Since any being not rational in this sense will not be under moral obligation, the criteria of rationality here must be satisfied not only by the perfectly rational but also by the imperfectly rational wills that Kant thought virtually all human beings to be.
The outline of the argument for negative freedom is as follows:
Implicitly the argument continues:
In order to avoid reducing the argument to obvious silliness, we need to guard against several temptations. First, the first premise cannot be read as saying that we “feel free.” Whether true or not, this would be a contingent empirical premise about us incompatible with Kant’s insistence on an a priori method; what is needed is rather a necessary truth about rational agency. It would not follow from the fact that one could not act without feeling free that one was free even for practical purposes; just as it would not follow from the fact that a professor could not lecture without feeling brilliant that he was brilliant for practical purposes. What is so “for practical purposes” is at least a reasonable assumption to make in deliberating and deciding what to do, and neither assumption is reasonable to act on just because it is an unavoidable feeling associated with the activity in question.
Second, the first premise should not be read as saying that those who sincerely believe in thorough-going determinism cannot act or, conversely, that no one who acts ever believes sincerely in thorough-going determinism. Surely many philosophers and scientists have believed in complete determinism, and it is too much to suppose that they lose that belief whenever they do anything. “Acting under the idea of freedom,” then, is more plausibly construed as “seeing oneself as free” or perhaps “taking oneself to be free” for certain purposes. The point of the first premise, then, would be that rational agents, in deliberating and deciding what to do, necessarily see themselves as free, regardless of their standing convictions on the metaphysical status of determinism. It is, one might say, a necessary condition of playing the game of deliberation.
Third, an important background of the argument is Kant’s belief, which he thought he had proved, that the idea of free agency is such that there can be no empirical evidence or sound metaphysical (“speculative”) argument that rational agents are unfree. If there were strong reasons for believing that rational agents are not free, then one could not so convincingly argue that it is reasonable to assume the opposite for practical purposes just because that is how rational agents must see themselves in acting. By analogy, suppose (implausibly) that in acting rational agents necessarily see themselves as indestructible Cartesian souls but that there are good empirical and/or philosophical arguments that this is an illusion. Then we would naturally be reluctant to conclude that for all practical purposes they are reasonable to assume that they are indestructible Cartesian souls, for practical purposes include deliberation about life-risking activities and their assumption has implications for this which they should reject no matter how they unavoidably “see themselves.”
Finally, we must not construe the conclusion that rational wills are free from a practical point of view as merely a repetition of the first premise that in acting rational wills necessarily see themselves as free. If construed that way, the argument would go nowhere. The conclusion is that for all purposes of deliberation and decision it is reasonable to accept all implications of the assumption that one is free and unreasonable to let one’s deliberations be influenced by the contrary idea that one is, or might be, unfree.22 We need here a distinction between merely seeing oneself as free, which is (by the first premise) unavoidable in rational deliberation, and taking full account in one’s deliberation of all the implications of the assumption that one is free. This is important because, by the rest of Kant’s argument, the implication of the assumption that rational wills are free is nothing less than that they are rationally bound to morality. The inevitability of rational agency (expressed in the first premise) is taking oneself to be free when deliberating and acting; the resulting prescription for all rational agents (expressed in the conclusion that we are free from a practical point of view) is that it is only rational to act on the full implications of the assumption that we are free. The latter is not inevitable but is what, according to the argument, we rationally should do.
Putting the pieces together, then, the argument runs as follows: Rational agents necessarily see themselves as (negatively) free when deliberating and acting; this means not only that they look upon themselves as choosing among options the outcome of which is not determined by prior empirical causes, but also that they see themselves as capable of reaching a decision in a way that is not a function of their given desires and their beliefs about the means to satisfy them. As in theoretical judgments guided by reason, rational agents deliberate with the view of themselves as able to reach reasoned conclusions which do not fit, or well serve, what they feel most inclined to. Given the impossibility of proof or even evidence that this view of themselves is illusory, they should accept, for all practical purposes, the idea that they are free in this sense. That is, they should accept any implication of the idea that “as a rational agent I am (negatively) free” as a reasonable assumption in all their deliberations about what to do.
This conclusion, we should note, is all that the remaining argument needs. For the argument as a whole is a practical one, addressed as it were to those deliberating about what it is rational to do and, in particular, to those wondering whether philosophical arguments concerning practical reason should be allowed to undermine their confidence that it is rational to be moral. To this audience the argument for negative freedom says, in effect, it is perfectly rational for your deliberative purposes to assume that, as a rational agent, you are negatively free: that is, you should assume that any account of practical reason is mistaken if it denies your ability to choose independently of determining causes or your ability to act for reasons other than desire-based hypothetical imperatives.
The preceding remarks are intended to reveal, or reconstruct, the initial argument for negative freedom as more coherent and plausible than it may at first appear; but two residual doubts should at least be raised. First, now that so much seems built into the first premise, what reason do we have to accept it? In particular, why suppose that rational agents must see themselves in deliberation as capable of being guided by rational standards other than maximum preference satisfaction? Kant believed that the latter standard could not be established as necessary by either the analytic or synthetic a priori methods he acknowledged; but has this been shown? Second, even if the sort of rational wills Kant had in mind necessarily see themselves as Kant says, why believe that we have rational wills of this sort? Perhaps we can form an idea of such rational agency but have to content ourselves with more mundane standards.
The argument we have been considering seems to rest the case on a thought experiment, or (more grandly) a phenomenological test: “Just try to see yourself (in acting) as lacking negative freedom and you will discover that you cannot.” But this reply, as Kant apparently realized, may be unconvincing by itself. Even if we discover what Kant expects, how do we know that the test reveals anything more than a universal but contingent feature of human nature? Also since the cases in which we are most convinced of our capacity to act rationally independently of given preferences are likely to be cases of duty versus inclination, might not our conviction be due to the fact that we have presupposed the rationality of moral conduct (which is what was to be established)? These worries would lead naturally to Kant’s discussion of the possibility of circular reasoning and his introduction of the “intelligible world” as the “third term” between reason and freedom. The latter idea, I suspect, stems not so much from obsessive concern with an otherworldly metaphysics as from the thought that even theoretical judgments, in science and everyday life, presuppose that we are guided (or guidable) by standards of rationality which, though applicable to experience, are not derived from experience and which are importantly different from “Find the conclusion that best suits your given preferences.” That we are the sort of beings capable of being guided by such standards is supposed to be evident not only in cases of moral conflict or other practical choices but also in theoretical judgments. The reply assumes Kant’s idea of the unity of theoretical and practical reason, which Kant does not try to defend in the Groundwork; and it raises deep questions beyond the immediate aim of this paper.
The preceding steps leave us with the striking conclusion that any rational agent is, as such, committed to at least some principle of conduct acknowledged as rational but not based on his desires and the imperative to take the necessary means to them. The underlying idea is that practical reason, like theoretical reason, enables us to reach conclusions on some basis other than that they get us what we most want. The next step, of course, is to identify the rational principles acknowledged by wills with autonomy as moral principles, the supreme moral principle and (perhaps) its derivatives. Kant attempts to do this in the first and second chapters of the Groundwork by (a) arguing that the supreme moral principle is “Act only on maxims which you can will as universal law” and (b) arguing through successive reformulations of this principle that it is “the principle of autonomy,” i.e., equivalent to saying, “follow the laws of a rational will with autonomy.”23
Though I believe these steps to be fundamentally flawed, I must postpone the attempt to argue the point and to explore a more promising route to a later formulation of the supreme moral principle. Nevertheless I will venture one general comment on this final stage of the argument. The main problem, I suspect, is that Kant switched illegitimately between two quite different readings of his famous first formulation of the supreme moral principle. The first reading is what naturally emerges from the argument we have been considering. Ask what principle, if any, must a rational will with autonomy accept, and an obvious, though rather unhelpful answer will be “Act in such a way that you conform to laws, or rational principles of conduct, you (or any rational being) accept independently of desire.” Assuming that all morally relevant acts can be construed in terms of their maxims, this can also be expressed “Act so that your maxims can be willed consistently with whatever rational constraints you (and others) are committed to as wills with autonomy.” Now in his transitions to the first formula of the supreme moral principle, Kant seems to be assuming that this is just what the formula says. What it declares, he says, is “conform to universal law as such,” where “universal law” has been defined to exclude rational considerations based on desires (G, 402, 420-21). So far, so good: that is, the formula is readily seen as one that any rational will with autonomy must accept. But now the trouble begins when Kant treats the same first formula as identical with, or as entailing, a principle that one must act only on maxims which one can will as universal laws in the sense that it is (rationally) acceptable that everyone act on the maxim. This moves from an undeniable formal principle to a dubious substantive principle; and despite all the brilliant aid Kant has received from sympathetic commentators, I fail to see how this transition can be made legitimately.
I conclude, however, with a more modest point. If I am right that Kant’s transition to a substantive supreme moral principle is illegitimate, this would be quite in line with my main reading of his argument. For my hypothesis has been that the task which Kant saw as most difficult and most sorely needed was not to convince anyone to accept his supreme moral principle or even to believe that if there are laws of autonomy they must be the familiar moral constraints. These matters, I suspect, he considered relatively easy, and perhaps for this reason he did not bother much about his argument for them. The harder task, and what was most needed, was what he thought he had accomplished in the earlier stages we have considered above: namely, to show that, despite philosophers’ arguments to the contrary, there must be principles of rational conduct other than desire-based hypothetical imperatives. That too may be an error, but Kant’s argument for it deserves more attention.
This paper is a compressed version of ideas presented at a conference on Kant’s ethics at Johns Hopkins in the summer of 1983, and an earlier version was also presented at the A.P.A., Pacific Division, meetings and at a meeting of the Triangle Ethics Group in Chapel Hill in the spring of 1985. Thanks are due to the participants at those meetings and also to Tyler Burge, Stephen Darwall, Gregory Kavka, William Lycan, Christopher Morris, and others for their helpful comments. A special thanks is due to Burge for constructive help and encouragement in long discussions on these matters.
There are exceptions, e.g., Dieter Henrich’s “Die Deduktion des Sittengesetzes” in Denken im Schatten des Nihilismus, ed. Alexander Schwan (Darmstadt: Wissenschaftliche Buchgesellschaft, 1975), pp. 55-112, in this volume pp. 299-337; Karl Ameriks’s Kant’s Theory of Mind, chap. 6; and Bruce Aune’s Kant’s Theory of Morals (Princeton: Princeton University Press, 1979). Two recent important books, published after this essay, are Henry Allison’s Kant’s Theory of Freedom (Cambridge: Cambridge University Press, 1990), and Rüdiger Bittner’s What Reason Demands, trans. Theodore Talbot (Cambridge: Cambridge University Press, 1989).
I assume that the person who seriously asks “Why be moral?” judges that he morally ought to do certain things but still has some sort of question. But on Kant’s view, to judge that one morally ought to do something is, in part, to believe that the conduct in question is required by an unconditional “command of reason”; and surely any (even minimally) rational person who believed this could not be indifferent to the conduct. This fits with Kant’s repeated suggestions that respect for moral law is, as it were, forced from us, that even murderers acknowledge the justice of their (death) sentences, etc.
Critique of Practical Reason, trans. L. W. Beck (New York: Liberal Arts Press, 1956), pp. 28–31[29–32].
H. J. Paton is an example; see The Categorical Imperative (London: Hutchinson, 1965), p. 221.
In Critique of Practical Reason, esp. pp. 29-31[30-31], Kant again argues for the mutual entailment of (a) rational beings are free and (b) rational beings are under moral law, but he says that consciousness of (b) is the grounds for knowing (a) and not the reverse. This is clearly a reversal of the order of argument in G, and probably represents an abandonment of that argument. But, whether Kant would want it or not, there does seem to be a way of reconciling the positions in the two works. That is, take the Groundwork as addressing doubts of agents deliberating about what to do, whereas the Critique is concerned more with completing a project on the nature of reason and which among traditional metaphysical beliefs one can reasonably accept (despite the arguments of the first Critique). The first, then, belongs to moral philosophy, traditionally conceived; the second to the special epistemological/ metaphysical/moral undertaking involved in the new critical philosophy. Since the audience of G is raising a practical question, it can be presumed willing to accept for those practical purposes anything implied by what it must assume in asking the question. Thus it need not demand any theoretical demonstration or intuition (“intellectual” or otherwise) of freedom if it is made clear that freedom is presupposed in its questioning. The audience of the Critique, on the other hand, may be thought more concerned with the justifiability of certain metaphysical beliefs, not from sheer speculative curiosity of course but also not from a practical need to satisfy the doubts about the rationality of morals (expressed at the end of the second chapter of G). If so, this audience is prepared to accept what the other audience was ready to entertain doubts about (namely, that as rational agents we are really under moral obligations). And, if the Critique’s audience is not seen as engaged primarily in deliberation about what it ought to do, the argument of G from the implications of deliberation is less appropriate. In any case, it is worth noting that the argument of G does not commit any of the mistakes in arguing for freedom that the Critique condemns (e.g., supposing there is an intuition of freedom).
This should be clear at least in my reconstruction of the argument, for none of the premises presuppose moral obligation. As Karl Ameriks helped me to see, the suspicion of circular reasoning may arise from worry that the first premise in the argument for negative freedom (i.e., rational wills cannot act except under the Idea of freedom) might seem convincing only because it is most evident in conflicts of duty and inclination. If so, the “solution” (which emphasizes that in all uses of reason we view ourselves as members of an “intelligible world”) would have a point without admitting that the original argument ever actually presupposed moral obligation. More needs to be said about the supposed circle and its solution, but this goes beyond the main argument I try to reconstruct here.
The argument reconstructed in this paper is an argument that morality is rational but it does not rely on the forbidden means-ends reasoning.
G, 448. This is the heart of (1) in the summary of my next paragraph.
G, 447. I assume that the strange formula in this passage, “the principle of acting on no maxim other than one which can have for its object itself as at the same time a universal law,” is meant to be the first formulation of the Categorical Imperative, and so have reformulated it.
G, 447. “This is precisely the formula of the Categorial Imperative and the principle of morality.” This is argued twice, at G, 402 and G, 420-21.
Kant clearly intended the principle to be unconditional and overriding, rather than prima facie, but one may question whether the argument supports this. In fact the sort of argument he gives seems unlikely to support an overriding principle; for the idea is to show that free rational agents unmotivated by desire must nevertheless have some rational principle to follow. Why would a principle that says what one prima facie ought to do not suffice? If this is all the argument supports, then the conclusion must be more modest, i.e., that every rational agent has some reason to be moral or immorality is always opposed to some prima facie principle of the agent. I will pass over this problem here, not only from lack of time, but also because the problem concerns later stages of the argument which are not my main concern.
The idea here is that there are two routes to show the rationality of moral conduct. One is to show that immorality conflicts with a commitment of the agent, whether or not that commitment was independently demanded by reason. The other is to show that rational agents, as such, necessarily are committed to a principle. I suspect that Kant intended both; for he argues that all minimally rational agents have autonomy and so are committed to the supreme moral principle, and he argues that necessarily, qua rational and free, they must accept the supreme moral principle. It is worth noting, though, that if the argument failed for the second, stronger claim, the first, weaker claim still would need to be considered. That strategy fits with the often repeated point that immorality involves conflict of will.
This account is most evident in R. P. Wolff’s The Autonomy of Reason (New York: Harper and Row, 1973).
The two views of will are discussed in John Silber’s introduction to Religion within the Limits of Reason Alone. Silber thinks that Kant was ambivalent at least in the Groundwork between the later view and the dualistic picture presented by Wolff. Beck’s commentary on the second Critique also makes much of the Wille vs. Willkür distinction. Rawls, I am told, distinguishes in lectures between the “Manichean” view and the “Augustinian” view of will, finding Kant ambivalent in the Groundwork. My concern here is not so much to deny the ambivalence, or confusion, as to see how the argument of the Groundwork does not rest on the more simplistic picture.
Strictly, the will is not even a hidden, unobservable event in time; to ascribe a will to a person is not to refer to a mysterious event or thing but merely to say, without further explanation, that the person has a capacity to make things happen in a way that makes appropriate the explanation “His reason . . . ,” “He was guided by the principle . . . ,” etc.
See the section entitled “Of Liberty and Necessity” in David Hume’s A Treatise of Human Nature, ed. L. A. Selby-Bigge (Oxford: Clarendon, 1955), bk. II, part III, pp. 399-412.
Kant suggests that it is because the will is a kind of cause that it cannot be “lawless,” but I take it that the point follows as well from the definition of will as acting in accord with the idea of principles or laws. There is no doubt a play on different sense of lawful and lawless, but the main point seems clear enough, namely, that to attribute an event to a person, thing, or prior happening as its cause (source, or author) we need some appropriate connection between the event and the alleged source, something to warrant saying the event occurred in some sense because of the source. Often the connection is an observable regularity between types of events, and then the “because” is an empirical causal one; but in the case of human action, on Kant’s view, the connection is between the event and a person’s beliefs and policy commitments, and then the “because” must be of a different type. This leaves much unexplained, of course, but the minimal point is just that some connection must be made, that this can be of two kinds, and that both involve “laws” in some sense or other.
See G, 443-44. Note that we have heteronomy as (mistakenly) regarded as the “basis of morality,” as a characterization of a rule or (would be) law, or as “the result” if one tries to make hypothetical imperatives the basis of morality (G, 441). It is significant that Kant’s discussion of heteronomy concerns not the ordinary following of prudential maxims and rules of skill but wrongheaded philosophical attempts to base morality in various ways on hypothetical imperatives.
I owe a thanks to Gregory Kavka for alerting me when, in an earlier version, I was yielding to this temptation.
This objection comes from Stephen Darwall, who may still be dissatisfied with my sketchy reply.
There is a sense, of course, in which even the non-conditional “Always pay your debts” prescribes options, for one may pay in coins or in bills, in cash or by check, etc. I trust, however, that patience could make clear how the relevant sense differs from this.
Strictly speaking, I suppose, the claim that, from a practical perspective, rational wills are free implies not only (a) that one should, for purposes of deliberation, accept all the implications of assuming that one is oneself free but also (b) that one should, for purposes of deliberation, accept the implications of assuming that others one takes to be rational are also free. The latter would be of practical importance when trying to decide whether execution for murder is justified. Unfortunately, the argument for (b) has to be more complicated and raises special problems. So I shall ignore it here. Implication (a) should suffice for most practical purposes, e.g., to dispel the doubts about one’s own rationality in being moral which are raised by either determinism or the thesis that all rationality is means-end calculation to maximum desire-satisfaction.
This culminates at G, 440.