a systematic failure in Kant's ethics

i

in the discussion of the thesis of the third antinomy, in the antinomy of pure reason, from the Critique of Pure Reason, Immanuel Kant says that the transcendental idea of freedom does not by any means constitute the whole content of the psychological concept of that name, which is mainly empirical. (critique of pure reason, a448) later, in the solution of the third antinomy, this specific transcendental idea is further described: freedom ... in its cosmological meaning [is] the power of beginning a state spontaneously. ( critique of pure reason, a533, emphasis in original.)

when it comes time for Kant to revisit the notion of freedom in his ethical writings, we find that it is the specifically transcendental—cosmological—sense in which freedom is at stake. freedom would then be the property this causality [i.e. the will] has of being able to work independently of determination by alien causes, just as natural necessity is a property characterizing the causality of all non-rational beings—the property of being determined to activity by the influence of alien causes. (groundwork of the metaphysic of morals, 446, emphasis in original.) the identity of freedom with rationality is later pressed home: it is not enough to ascribe freedom to our will, on whatever ground, unless we have sufficient reason for attributing the same freedom to all rational beings as well. (groundwork, 447)

Kant supposes his idea of freedom here, in a move which he borrowed from the critique of practical reason, to be founded in a fundamental dualism within human beings—or rather, within all rational beings. this dualism has it that human beings, considered as objects of experience, are determined by the same network of physical causes as all other beings, but that when considered as things in themselves, they might have true (transcendental, cosmological) freedom as well. what was left as a possibility in the critique of pure reason is relied upon heavily in the groundwork of the metaphysic of morals, and in the critique of practical reason as well. the point is made most forcefully in the metaphysics of morals:

the concept of freedom is a pure rational concept, which for this very reason is transcendent for theoretical philosophy, that is, it is a concept such that no instance corresponding to it can be given in any possible experience, and of an object of which we cannot obtain any theoretical cognition. (the metaphysics of morals, 221, emphasis in original.)

at the outset, however, it is necessary to address a preliminary point. i argue in this paper that there is a fundamental difficulty in connecting Kant's ethics with his metaphysics. (some would say contradiction, but this is not, strictly speaking, true. my method here is to work, taking it for granted that there is no contradiction, and show that the result is that the ethics is nearly empty. one could, instead, argue that the emptiness demonstrates that the ethics must not work in this way, and then that there is a contradiction between the metaphysics and the ethics. i am agnostic on the relative value of these two approaches; i choose the one i follow here because i find it simpler to explain.) if it rested there, it would be a relatively minor ad hominem point. however, the metaphysics of Kant is not merely some additional beliefs he happens to hold, which one might freely jettison and still think oneself to be an adherent of his ethical system. Kant himself believes his metaphysics to be directly necessary in making sense of the ethics.

the first two chapters of the groundwork are concerned with drawing forth the shape and consequences of a categorical imperative if there happens to be one. finally that conditional is resolved in the third chapter, in which Kant argues that we must presuppose freedom as a property of the will—even if we do not, the same laws as would bind a being who was really free are equally valid for a being who cannot act except under the idea of his own freedom. (groundwork, 448n.) in this way, as noted above, it is not necessary for Kant to argue for the existence of freedom simpliciter, provided he can describe how we might act under the idea of freedom, which is to say, as if we were free, even if we were actually not.

so, human beings are not necessarily free in the last analysis, but rather, agents must take themselves to be free in their action, and this is, for all strictly practical purposes supposed to be just as good as having an actual metaphysical proof of freedom. (in the kantian notion of a practical purpose, of course.) alas, however, it is not just as good. for, the issue of freedom (or rationality, which has been inextricably linked to it) does not arise only when considering one's own status as an agent, but also, as the formula of humanity makes clear, when considering all other objects of experience as well.

ii

consider a typical situation in which a kantian duty applies. suppose for example you are a doctor with a patient in need of a heart transplant. enters into your office a twenty-five year old man, possessed of a healthy beating heart; this man will be undergoing surgery under general anesthesia to remove a bone-spur in his ankle. for irrelevant emotional reasons, you are strongly attached to the heart patient, but you dislike the ankle patient intensely. you might choose to do the following: when the ankle patient is under anesthesia, remove his heart and transplant it into your heart patient, thus killing the ankle patient and saving the life of the heart patient.

Kant of course would agree with most of us that such an action is morally reprehensible: it is contrary to a strict duty. it could be described under any of the forms of the categorical imperative, but for my present purpose the formula of humanity will do. always act in such a way that you always treat [rational natures], whether in your own person or in the person of any other, never simply as a means, but always at the same time as an end. (groundwork, 429.) the ankle patient is a person, possessed of inherent dignity. to remove his heart is to treat him as a mere means (to the end of saving the life of the heart patient), and is therefore prohibited.

but suppose alternatively we have an artificial heart available; suppose that medical technology has developed this remarkable device as a standard treatment, and it is able to prolong life better than our current technology happens to permit. could our surgeon take the artificial heart off the shelf and use it to help his heart patient? of course. but is not the surgeon using the artificial heart as a mere means? he is allowed to, under the formula of humanity, because indeed the artificial heart is a thing. things have value, while persons have dignity. it is allowed (indeed, probably required) to use available things in this way as means to the extension of life of persons.

consider then these two alternative scenarios. in both, the surgeon has a heart patient who will be helped by the surgical implantation of a blood-pumping device. in one case, the device comes from a box in the storeroom. in the other, it comes from a living patient expecting ankle surgery. whether our surgeon is allowed to implant the blood-pumping device depends precisely on whether the source of that device is a thing or a living person possessed of a rational nature. simply put, to give content to the formula of humanity, one must know which objects in the world to take as having rational natures. the same issue arises under the other formulations of the categorical imperative. applying the formula of the kingdom of ends requires that we know which entities are citizens of that kingdom; applying the formula of universal law requires that one know to which beings in the world the contemplated universal law applies.

iii

one immediate attempt at answering this question would be to find indicia of rational nature. for Kant there indeed are things rational natures can do which other things cannot. perhaps those properties will serve to answer the question. suppose then we had a list of indicia for rational natures. if they are to be of any practical help, they would need to be phenomenal signs which signal the presence of a rational nature. perhaps we might look to linguistic competence, neurological complexity, or some other such mark or combination of marks.

for Kant, the only morally relevant attribute of a rational nature is that it is capable of acting from the moral law. it is not merely enough that the being never happens to violate a duty. indeed, given that duties track capacities (you don't have a duty to do something if it is a strictly impossible thing for you to do), it would be easy to suppose that stones indeed never violate their duty. Kant is aware of this—the attribute of rational nature is not merely acting in accordance with duty, but acting from the motive of duty. (of course, Kant would think it makes no sense to speak of a mere thing as having duties at all. i use that language to mean rather the duties that thing would have, if it were to have a rational nature.)

motives are hard to get a handle on, however. indeed, Kant says that even the acting subject is not aware of his own motives in this regard (groundwork, 406–409). if we want to find indicia of rational natures, they must be indicia that live in the phenomenal realm, indicia which we can observe. certainly there are differences between beings which possess a rational nature and those which don't—that is not the issue here. the issue here is how one can tell, of the various beings in the world, which have a rational nature and which do not.

consider a proposed indicium of rational nature, say some kind of linguistic competence test. linguistic competence as a phenomenal can always be explained satisfactorily by cause and effect, purely within the realm of phenomena. the thing-in-itself does have a noumenal reality, and in that noumenal realm it might, or might not, have reverence for law as part of the explanation for its linguistic acts. but we cannot tell. if linguistic competence were a sign of a rational nature, then rational natures would be phenomenal realities: for that is precisely what a phenomenal thing is.

any list of suitable indicia indeed must be unacceptable to Kant. it is not important whether the indicia are perfect or not. if we had any list of indicia whatsoever, then the wall between phenomena and noumena would be abolished. there can be no phenomenal signs of rational natures, as long as rational natures are required to be purely noumenal properties, which they must be to play the critical role of allowing for freedom in a world of causally-determined phenomena.

iv

the previous section describes why Kant cannot have epistemological grounds for ascribing rational natures to any particular class of objects in the world. but perhaps we do not actually need knowledge of which things have rational natures. Christine Korsgaard suggests, with reference to responsibility and its allied notions of praise and blame, that we do not know when to ascribe responsibility, precisely because we do not have access to the motivations of another. but she sees this as a flaw in assuming that responsibility is some kind of attribute of the other, which they might or might not have. instead Korsgaard articulates a different way to think about the assignment of responsibility: it is a choice we make, not a description of a thing out there, but rather a difference in how we interact with the thing. (creating the kingdom of ends, chapter 7, 188—221. see especially 189—197.)

can this work to solve the problem of identifying rational natures? this kind of strategy would suggest that we not think of the problem as one of identifying rational natures, but instead of how we choose to interact with beings in the world. We can interact with them in a rational-nature-ascribing way, or in a purely-just-a-thing-ascribing way. we would no longer need indicia of rational nature. we would still need some idea of when it would be best to adopt one stance and when to adopt the other stance. before tackling that problem, this strategy does have some initial successes.

first, consider punishment. for Kant, the punishment of a malefactor is a matter of obligation: you in fact owe it to the malefactor that he be punished—but only if the malefactor indeed has a rational nature—it's no good punishing stones. consider then a malefactor, and suppose all doubt of guilt is removed. should we punish? if the malefactor has a rational nature, yes. if not, no. or wait—if the malefactor has no rational nature, then indeed we may not be obliged to punish, but surely we are permitted. we can freely jail stones, for example. (there might still be some residual duties here, akin to the indirect duty not to torture animals.)

so this strategy of standpoint choice yields a desirable result in accounting for punishment. indeed, we do not need to know whether the being has a rational nature or not; we simply jail the criminal. if he has a rational nature, then we are doing to him what we owe him. if not, then there is no person there to harm at all.

we can also ground duties to the self. insofar as one acts and chooses which options to enact, one ipso facto adopts the stance that one's self is a rational nature. accordingly, duties to the self can be grounded in just this way. i cannot simultaneously choose what to do and adopt a stance under which i am not possessed of a rational nature.

flush with this initial success, one might hope to expand this kind of treatment more broadly. but it does not in fact go much further that what i have just sketched. indeed, it has real difficulties even going that far. consider the case of punishment. the analysis i give explains why punishment still makes sense even if we lack epistemological grounds for ascribing rational natures. but which beings do we subject to this analysis? this might explain how we can reasonably punish humans, but it does not yet explain which beings we should punish, given that in fact there are some we never punish at all.

suppose a tree branch falls on a pedestrian, killing him. should we punish the tree? if we truly claim epistemological ignorance of the noumenal being of the tree, then we do not know whether the tree has a rational nature or not. according to the analysis above, we should punish the tree: either we are doing to the tree what is owed for its malfeasance, or we are just chopping down a mere thing we have liberty to use as we please.

it might be argued that this kind of standpoint-choice is only intended to be used when we are in doubt, and we are not really in doubt about trees. what that really means, however, is that the tree, lacking such things as complex neurology and linguistic competence, is simply being assumed to lack a rational nature. the basis of this common-sense assumption is in attention to properties which, for Kant, must be strictly irrelevant. if we only need to apply the standpoint-choice approach in doubtful cases, then it could be of great help. cases which are not the least doubtful would not require the standpoint choice approach.

unfortunately for this approach, the argument in section iii above is not about borderline cases at all. it is not merely that we don't have good indicia to distinguish in the hard borderline cases between possession of rational natures and lack of them: it is that we lack any indicia whatsoever to distinguish stones from people, as regards rational natures.

there is one final gasp available to the standpoint-choice approach; suppose that since one does not have moral duties to things, one will simply act conservatively, taking all things as potential bearers of rational natures. if one does this with strict equality, then kantian ethics would become contentless; we would in fact be treating all beings as persons, and be unable to act at all. (our surgeon from section ii would not even be able to use the artificial heart to save a life, because that artificial heart might have a rational nature.)

one might suggest some kind weighting of likelihoods though; perhaps we might say the stone is very, very unlikely to have a rational nature, the tree merely very unlikely, the flatworm unlikely, the ankle patient likely, and the erudite philosopher very likely. on what basis do we make this determination however? it can only be because we know which properties make one more likely to have a rational nature: because we have some kind of imperfect indicia of noumenal realities.

v

we could simply assign this the status of being a mere philosophical puzzle, something we can file away and say it is future work to be done. however, there are many live ethical issues which depend directly on how we attempt to figure out to which beings we have duties. Kant assumes we have duties to human beings, on the unproven and unargued grounds that they are persons, and not to rocks, on the unproven and unargued grounds that they are not. we should expect from an ethical theory some kind of guidance on how to decide to which beings we have duties and which we do not, and given that there are open issues in this regard, kantian ethics falls short.

for example, Kant is quite clear that non-human animals are not persons. at best, his argument for this position is that of Descartes: they are simply automata. but if that were to settle the question, then it should settle it for humans as well; Kant thinks that all phenomenal reality is tied by unbending natural laws of cause and effect.

again, not all humans are possessed of equivalent psychological ability. there are children, infants, the mentally ill, anencephalitic newborns, unborn embryos and fetuses, and a host of other varieties of human existence. but what guidance can Kant give about how to decide whether we have duties towards such beings? there are no phenomenal signs that will do: nothing in their appearance—nothing—can tell us whether they have rational natures at all.

(indeed there is a related problem here that is too far afield to get a serious treatment in this paper: what about the possibility of varying degrees of rational nature? Kant takes it to be a one-size-fits-all sort of thing, but is this realistic? we might be more inclined to regard it as a more flexible and complex reality than just a binary yes/no question, does this being have a rational nature or not? not only does Kant fail to give a satisfactory answer, he fails even to give us the terms in which to understand and evaluate possible answers, because nothing about the beings in question can tell us whether they have a rational nature.)

interlude

one might think that the question do stones have rational natures? is so ludicrous as to be beside the point of serious ethics: but that is because we all operate, it seems, with some background assumptions about rational natures. stones do not participate in social interactions, have no linguistic capacity, lack any kind of appropriate neurological complexity: we just know, we say, that they lack rational natures, and we resent demands for proof.

but we cannot accept such reasoning on monday, and then on tuesday turn around and say that rational natures are noumenal realities, permanently excluded from phenomenal observation. if the question why do we know stones lack rational natures? is a trivially simple one, not worthy of serious consideration, then it might be annoying to be forced to provide an answer, but there had better be one.

alternatively, one might wish to adopt Kant's ethics, and not his metaphysics, with its attendant restrictions on our knowledge. can Kant's ethics survive such a truncation? it seems to me that it cannot. the thesis that freedom is something we can never directly see in ourselves, let alone others, is a cornerstone not just in the relevant sections of the critique of Pure Reason, but also in the second and third chapters of the groundwork of the metaphysic of morals, the introductory portions of both parts of the metaphysics of morals and throughout the entire critique of practical reason (consider, for example, how one should understand the discussion that we cannot really know for ourselves whether we are truly acting from duty alone. can this discussion even make sense if it is cut off from the kantian notion of a hiddenness of freedom?)

in the remainder of this paper i consider several ways one might attempt to sidestep the need for a method for discriminating between rational natures and mere things. i attempt to consider seriously first the possibility that we might take a conservative approach. then i will consider two methods of avoiding the question entirely. finally, i will suggest some possible moves for ethics in general, and explain why, however pleasing those moves might be, they are fundamentally non-kantian.

vi

in general the kantian response to practical doubt about duties is two-fold. first, one should engage in casuistical analysis of the doubt: turn it over, mull on it, compare it with other cases, and so help to figure it out. this kind of process is not imbued with any kind of certainty, but perhaps that is actually desirable in a process that helps us deal with doubtful cases. however, as i have described above, no casuistical analysis of this particular sort of doubt can proceed, because we lack any understanding of even the broad outlines. that is, we cannot discuss the case of animals and infants by analogizing to various relevant features of normal adult human beings, because we have no account of what those features would be, and indeed, there is a systemic obligation for Kant to deny that there are any such phenomenal features. normally casuistical analysis proceeds by comparing the doubtful case to ones more certain, but we have, in this situation, no cases which are any more certain.

the second kantian method of dealing with such uncertainty is conservatism. if we are doubtful whether a duty obtains in some context, we can always simply act as if it did. staying thus always on the right side of the law (building a fence around it?), we manage to fulfill our duty even when there is doubt. this kind of conservative response, however, is inadequate to the task. as suggested above in section iv, conflicts of such potential duties arise constantly. to play tennis, one must ignore the possibility that the tennis ball has a rational nature. this might seem flippant, but again, we have absolutely no indicia of rational natures.

in other ethical frameworks besides Kant's, it works perfectly well to say that one is conservative in borderline cases (infants, say) but as one moves further from the fuzzy boundary (tennis balls) one worries less and less about the possibility that one is transgressing a duty to some undetected rational nature. those other ethical frameworks can at least provide a description of the rough outlines of this boundary. for Kant, it is strictly impossible to provide any description of the boundary at all. so the conservative attempt to dodge the bullet fails. it remains to be seen whether the question can be avoided entirely, and then to consider what remains for ethics in general after the entire discussion.

vii

however, it does not in fact have the requisite structure. it presumes that we know ourselves to be bound by law, we feel it in our bones, so to speak. all laws are conditional on the existence of the conditions for the law's application (the law against lying upon the existence of speech, the law against killing upon the existence of a means of killing, and so forth). without those conditions, the law still applies, just as strongly, even though it has no practical effect in the circumstances which obtain. the attempt to build an argument that proves the existence of those conditions must fail: we can know ourselves to be bound by a law in this way without knowing that the conditions for that law's application actually obtain. (for example, the application of the duty to care for one's children is conditional on the existence of the children. if you don't have children, then you still have the duty to care for whatever children you have, and you know yourself to be bound by that duty, even though the conditions for its actual application happen not to obtain. just so, that we supposedly might know ourselves to be bound by duties to other rational natures, should they happen to exist, does not somehow establish the existence of those rational natures.)

alternatively, one might not accept the premise that we feel morality in our bones. in that case the attempted argument fails even more quickly. far from demonstrating the necessity of assuming the existence of rational natures, it would instead simply point out that we must be as ignorant of the existence of morality as we are of the existence of other rational natures. and for Kant, the argument cannot even really get that far. kantian duties are not all towards other beings, but include duties to ourselves. if there must be at least one real duty in force for morality to exist, this still does not require there to be at least one real other rational nature. duties to self will fill the bill quite nicely.

viii

what exactly is the class to be? i must take myself to be a member of it, in considering my own actions, at least. should the bounds of the class be my species? or linguistic competence similar to mine? Kant seems to have been a firm believer in natural categories of such things. so give him the existence of natural kinds. it does not follow at all that the natural kinds must have anything to do with noumenal realities.

indeed, natural kinds are precisely the sorts of things that Kant believes are strictly phenomenal. they are described by cause and effect relations; that which establishes their kinship is precisely their character vis-a-vis other natural things, all of which takes place in the world of experience, the world of phenomenal reality.

moreover, there are fuzzy boundaries around the obvious candidate class (human beings). Kant himself famously did not much discuss the case of children as equal ethical subjects or objects nor other such cases of apparent partial rational nature and subrational human beings. any attempt to sidestep the question which beings have rational natures? by asserting that there is only one candidate class of such beings to consider would require that there really be only one candidate class. in fact, there are several, and an important problem in ethics is precisely the task of sorting them out.

ix

the problem outlined here, of identifying the proper objects of moral concern in the world, arises for any ethical theory, whether kantian or not. i will now briefly describe three methods which one might use for addressing the problem of other minds in ethical theory, and in each case explain why it will not suit for Kant.

a. first, one might simply bite the Korsgaard bullet fully, and say that we have duties to other beings only when we do in fact choose to view them as rational natures. Korsgaard only applies this to questions of responsibility and allied notions. but what happens if we use it more broadly?

the first obstacle would be to explain how and when we should extend such moral consideration. at first blush, it would appear that we could simply choose never to do so, and then live in carefree abandon of others. but we must be truthful about the stance we adopt to others; we cannot honestly view them as having rational natures for one purpose, while simultaneously claiming not to in order to step on them to get what we want.

typical methods for extending consideration therefore might be social pressure, emotional attachment, and the like. ideally, we could have ethical conversations about whether we should extend consideration to some class of beings heretofore ignored. such a conversation would treat of the kinds of things identified above in section iii as indicia, and perhaps additional things like such as emotional and social factors.

but at no point would we be permitted to claim any kind of ontological correctness about what we have done. there would be nothing correct or incorrect about choosing whether to extend moral consideration to a particular being; and so there is no right or wrong about it either. all we can say is that if we have extended consideration for non-ethical purposes (if we do in fact view them as having rational natures), then we must recognize that a set of moral bounds goes along with it.

this is an appealing approach, but it is not available to Kant. these moral bounds look an awful lot like hypothetical imperatives (if you want to view x as a rational being, then you must behave in such-and-such a way). even if they are not actually hypothetical imperatives, they are still entirely subjective. kantian duties are supposed to apply to all suitably situated agents, and not merely to those who have chosen to engage in arrangements of reciprocity—or (God forbid!) those who happen to have certain kinds of emotional attachments. this approach, however attractive, will not suffice to ground the objective character a kantian duty is supposed to have.

b. a second strategy would simply be to adopt a skeptical solipsism. this is really quite the inverse of the conservative approach. instead of viewing all beings as potentially possessed of rational natures, the solipsist takes none of them to be, except herself, which she cannot help but take to be a rational agent when actually acting.

the ethical solipsist would be a person who takes the standpoint-choice argument of the previous subsection a seriously, but then decides carefully not to extend consideration to any beings outside herself at all. seen thus it's clear what's wrong with the solipsist: she isn't even a good egoist. such a person will be cut off from almost all the greatest sources of happiness and joy.

it seems clear that this approach of never extending consideration to others, of being as stingy as possible with applying duty to ourselves, might well be a consistent approach, consistent indeed as well with kant's program. but why indeed would one ever choose it? presumably the motive for choosing such an approach would be the usual strategy of an immoralist: to be free to do whatever one wants without any bounds of morality. but while our ethical solipsist may be freed from the bounds of morality, the price is simply too high to pay, and we might well object that in fact nobody is actually willing to pay it.

c. a final approach might be called straightforward realism. the realist (as i use the term here) simply says that such things as future-directed planning, linguistic competence, neurological complexity, species membership, and so forth, just are what a rational nature is. a complete list of necessary and sufficient conditions is not required as long as we have some idea, and the right kind of family resemblance still holds. we then evaluate where to extend consideration on the basis of these traits. there is a gray area (as we should hope for, for a successful theory) where some of the traits obtain but not others, or obtain only in partial degree.

such a realist approach may have little trouble explaining the subjects of this paper—borderline cases, justifying the list of criteria, and so forth. and, unlike the choice-of-stance approach, this strategy would hold that there are objectively correct criteria for identifying rational natures, and that we have a partial working knowledge of those criteria; one that we seek to improve over time.

the critical aspect of rational natures for Kant is that of being capable of acting from a motive of duty. and rational natures defined by a straightforwardly realist method can simply include such a capacity as one very important trait. where it is not visible, or we are uncertain, we might legitimately infer it from other things (such as future-planning and neurological complexity).

but for Kant, this trait is the only one that matters for morality, and it cannot be visible. this approach cannot be permitted to follow through for Kant—the entire argument of section iii of the groundwork would collapse. we come back then to where we began: any attempt to give phenomenal indicia of rational nature would subvert the noumenal character of that nature. indeed, rational natures on the straightforwardly realist account might have some kind of reference to moral law, but not of the requisite kind for Kant; not the kind that can establish the very possibility of the categorical imperative itself.

conclusion

acknowledgment

references