APPLYING THE CATEGORICAL IMPERATIVE(1)
There are many possible styles of doing moral philosophy, distinguished from each other in a variety of ways. Philosophers may differ about what the fundamental objects of moral judgement are, as between persons, institutions, states of character, actions, and so on. They may disagree, even, about what moral judgements are, ranging from a view that what are taken to be the verbal expressions of such judgements belong to a species of non-cognitive discourse (whose logic is the object of analysis) to a view that such judgements fall under the general category of judgements about the health of an organism. Doubtless these varieties, and others, are familiar.
An important tradition in ethics sees as one of the central tasks of the moral philosopher the discovery and elucidation of a single fundamental principle of morality. Characteristically, such a principle is one applicable to actions, though objects of judgement other than actions may be taken as fundamental (for example, in the case of teleological principles consequences would be the primary objects of evaluation). A philosopher who approaches ethics in this way faces a dual responsibility: he must state the principle he takes to be at the root of morality, and he must show how it is to be applied, and show schematically what the results of that application are. To apply a principle is to derive from it and a description of a set of facts a judgement concerning a particular action that it is, say, wrong. To give a full account of the application of a principle is to do this for all possible relevant descriptions of sets of facts. Even with the economy made possible by the principle's internal criteria of relevance, to give a full account would obviously, be an impossible enterprise; the most that can be demanded is an account which covers, schematically, a range of cases intuitively thought to be important.
An account of the application of a principle is no mere heuristic adjunct to the proper business of the moral philosopher. It is crucial. For one thing, the very sense of the principle is at stake: if (P.Q) entails R, and (P'.Q) does not, then P and P' are different principles, whatever their verbal formulation. But further, reflection on the point of doing moral philosophy reveals that a principle sans casuistical commentary is well-nigh worthless. The aims in question are twofold: On the one hand, we want a principle to bring a theoretical order into our judgement-practice, leading us to a discursive understanding of it, enabling us to see it whole. On the other hand, as with other theories designed to unify a practice of judgement, we hope that the principle will help us to greater lucidity in our practice, that it will lead us to expunge those prejudices and irrelevancies which lead our judgement askew. Neither of these aims can be achieved without a codicil explaining how the principle is to be interpreted. We would not see the unity in our practice if we could not trace the fabric of derivations along which our judgements flowed from the principle. And we could not purify our practice if we could not conduct those derivations for ourselves.
Philosophers have not shirked the first responsibility, that of putting forward candidates for the position of fundamental moral principle. But they have not always been so forthcoming with the second. Even a great philosopher may exhibit great ingenuity in arriving at the first principle he offers, but fail us hopelessly when it comes to instructions for is application. It is so, I think, with Kant. He leaves us no less thanfive allegedly equivalent formulations of his fundamental principle, but his work on the application of that principle is careless, prejudiced, and morally blind. To say this is not to demean the greatness of his insights into morality, or to diminish this stature in the history of philosophy; it is to point to an area in which we lesser lights may limn a surer path. At any rate, what I want to explore here is a path which avoids, I hope, the more miasmic of the Kantian swamps.
The basic formula of Kant's principle, the Categorical Imperative, is "Act only on that maxim through which you can at the same time will that it should become a universal law".(2) This, is the first of the five formula, is the one Kant recommends as a guide to moral judgement; his view of the others, to put an uncharitable construction on it, is that they are useful aids to moral propaganda. Kant himself is not entirely true to this recommendation; indeed I think that his deviation from it is symptomatic of part of what is wrong with his application of the Categorical Imperative. His first set of examples follows the introduction of the second formula, "Act as if the maxim of your action were to become through your will a universal law of nature".(3) The shift from "universal law" to "universal law of nature" may seem a small one; but it makes possible a nice piece of sophistry in the very first example, which purports to derive from the Categorical Imperative a perfect duty not to commit suicide. The maxim involved in the case Kant discusses is "From self-love I make it my principle to shorten my life if its continuance threatens more evil than it promises pleasure". Such a maxim is shown, without much ado, to be incapable of existing as a universal law of nature: "It is then seen at once thata system of nature by whose law the very same feeling whose function is to stimulate the furtherance of life should actually destroy life would contradict itself and consequently could not subsist as a system of nature".(4)
For pulling our eyes, this is poorly knit wool. Even granting the gratuitous assumption about the function of self-love (for which, perhaps, a Darwinian case could be made), the argument limps. Kant seems to be asking "Would you will a law of nature which said that an unfailing propensity to preserve life sometimes be used to end that life?". To this I am inclined to reply: "Well, no - not if you put it that way. But if Nature wants unfailing propensities, she'd better make shift for herself - I won't will them for her. A teleological law had better have its own licence; I won't make an honest proposition of it". To use the notion of function - a legitimate enough notion in its place - as Kant here uses it is quite unacceptable. After all, the function of the mouth is presumably to eat and talk; is it then immoral therein to store a mouthful of nails while one's hands are occupied with hammer and board?
Other cases in Kant's ethical writings where his application of the Imperative leads to unacceptable results are not hard to find. His fanaticism about capital punishment, and his antediluvian attitude toward revolution, are well-known. I find, however, his discussion of maternal infanticide the most arresting of all cases where he brings the modern reader up short (save perhaps for the more wrathful passages of Religion within the Bounds of Reason Alone). The type of case he discusses is that in which a woman who has given birth to a child out of wedlock murders (if I may use the word, pace Kant) the infant. In Kant's argument that such a killing is not subject to penal sanction, the following striking passage occurs:
A child born into the world outside marriage is outside the law (for this is implied by the concept of marriage), and consequently it is also outside the protection of the law. The child has crept surreptitiously in to the commonwealth (much like prohibited wares), so that its existence as well as its destruction can be ignored (because by right it ought not to have come into existence in this way); and the mother's disgrace if the illegitimate birth becomes known cannot be wiped out by any official decree.(5)
A year-round open season on illegal immigrants is presumably the next step! In fairness, I should note that the setting of this passage is a discussion of crimes arising out of a sense of honour (maternal infanticide and killing in a duel between soldiers). Preserving the appropriate type of honour, Kant says, is a duty for women and soldiers; since there is no way the state can preserve the honour of an unmarried mother or an insulted soldier, each is in a state of nature with the offender. Of course this view does not stand up to the minimal scrutiny involved in stating it: to treat as relevantly similar an innocent child and the author of an unretracted insult is morally grotesque.
These passages, and others like them, in Kant's ethical writings could lead us to one of two conclusions: either Kant's theory is unsound, or he is sometimes strikingly inept at applying it. In some ways the first conclusion would be more comfortable; it would allow us to treat Kant as of purely historical interest, and to take up an undivided attitude toward him, seeing his entire moral philosophy as hidebound and petty. But it is the second conclusion which must stand.
Whether or not Kant's theory is sound, the examples I have mentioned, and others like them, are not what shows it to be unsound. Kant's arguments about these examples would be silly whatever principle he were espousing, so if his principle is inadequate to found morality, it is not so because he reaches silly conclusions about the examples. His argument about suicide, for example, should not lead us to suppose that the Categorical Imperative provides an inadequate test for rightness (I am assuming that suicide is not unqualifiedly wrong), because Kant's attempt to show that there is a contradiction in the supposition that the maxim in question be universalized is an embarrassing failure.
To say all this does not inspire, necessarily, much optimism about proving Kant right. We cannot glean the sense of the Categorical Imperative simply from the words of the first formula (and certainly not from those of the others); we need some guidance from Kant about what, for example, he means by the impossibility of a maxim's being universal. One of the ways in which we might obtain such guidance is by following him through examples of the application of his test. But in a number of cases he is just the leader guaranteed to take us headlong over the cliffs of fallacy. So the very thing we need to get a clear idea of the meaning of the Categorical Imperative, textual evidence, is something our confidence in which Kant has grievously undermined. How are we to distinguish those cases where Kant is applying his principle roughly correctly from those in which he is torturing reason to extract a conclusion which fits his prejudices?
The problem is, of course, not as insuperable as it sounds. After all, we can be shameless, and follow Kant as it suits us, and not when it doesn't; exegesis of his thought does not require us to defer to his words blindly, line by line. And two general strategies recommend themselves for digging through the dross to the real stuff of his thought. First, the more general and the more abstract a passage, the more we can trust it. Where Kant is not to be trusted is where he is deriving fairly specific duties from the Imperative; it is there that psychological factors can knock one's judgement askew, that inherited irrational prohibitions can so strongly demand papers of legitimacy that one blinds oneself to forgery. Where Kant is not dealing with specific duties, where he is discussing in general terms the application of the Imperative, there we may take him, provisionally, to be objective. The second strategy has less to do with Kantian exegesis; it derives from what we know about principles which bear some affinity to the Categorical Imperative. These are what may be called generalization principles: they take as a crucially relevant consideration for judging the rightness or wrongness of an action the answer to the question "What if everyone did that?"(6)
These principles characteristically assess the rightness or wrongness of an action by reference to the consequences of everyone's performing an action of that type. A principle might say that if the consequences of everyone doing A would be undesirable, it would be wrong for anyone to do A, or, perhaps, that if the consequences of everyone's doing A would be desirable, it would be right for anyone to do A. No doubt the family is well-known. Kant's principle is not truth a member of it, though it may have surface similarities with paradigm members; the similarity can be exaggerated if one is careless about the formulation of maxims, as, for example, is Frankena,(7) who cites as a possible maxim "Tie your left shoelace first" - if such prescriptions counted as maxims, presumably the universal law test would be very like the "What if everybody ... ?" test. I hope it will become clear that this is the wrong way of looking at the universal law test.
The importance of generalization principles in this context is the manner in which they fail. Any principle analogous to paradigm generalization principles must meet the problem which makes those paradigms worthless; Kant's principle is, I think, sufficiently analogous to require an explicit account of how the problem is met. So keeping the problem clearly before one should be an aid to sympathetic interpretation - one should be looking for an interpretation which distinguishes Kant's principle from the generalization principles. The problem that has to be met is the problem of relevant descriptions. Applying a generalization principle requires one to answer the question "Would the consequences of everyone's doing this type of action be undesirable?" To this, the short answer is "Yes and no": types and descriptions can be paired, so we may say that the question concerns actions sharing the same description as a given act; but the act has innumerable descriptions, under some of which universal performance would have undesirable consequences, under others, desirable. Contradictory normative judgements naturally emerge from the principle: an action may be right if described as "Eating one evening meal a week in a restaurant", hideously wrong if described as "Eating in Giovanni's at 7:00 p.m. on Tuesday". Similarly, with appropriate choice of descriptions, all the alternatives open to one may turn out wrong; choose "Becoming a doctor" and "becoming something other than a doctor". If Kant's principle is to avoid this kind of absurdity, it must supply internally a non-arbitrary criterion for relevance of description.
I will attempt an account of the way in which the Categorical Imperative is to be applied which will, I hope, be true to the essential Kant - Kant the moral theoretician - and which will avoid the paradoxes of generalization principles. I will avoid too tedious a shuffle through the texts, since the account I will give will justify itself as a whole, or not at all. In the main, I will be working toward an account of morality that stands a chance of being correct, so far as that can be done without gross abuse of Kant's words. My procedure will be, first, to outline a first approximation of the manner in which the Imperative is to be applied, second, to discuss the way in which Kant treats the "false promise" case, and to examine whether the mode of application introduced deals with the problem of relevant descriptions, third, to modify the first approximation to meet the problems so far discussed, and fourth, to sketch the way in which sample effect duties to others may be derived from the Categorical Imperative.
The first approximation which I put forward, I should stress, is just that. Some of what I say now about the application of the imperative will be taken back later on. But the general schematic outline presented here is necessary for understanding of the later material. Initially, then, I take the derivation of duties from the formula of universal law to proceed as follows:(8)
1. Set out the maxim, M, on which the agent proposes to act, or has acted. M must be the actual maxim of the agent, not just some possible maxim bearing upon the same action. A maxim is a conditional intention designed to achieve an end; it could be represented schematically as "For the sake of end E, when in circumstances C I'll do A". M should satisfy the condition that if the agent's causal beliefs are true, doing A in C is a way of getting E.
2. Ask: Could a rational agent will M to be a universal law for all rational agents? (For present purposes this could be read as "... will that rational agents adopt M?"). If the answer is affirmative, M passes the test; if negative, then either:
(a) It is inconceivable, or self-contradictory, to suppose that M be a universal law for all rational agents.
(b) It is conceivable that M be a universal law, but no rational agent would will M to be a universal law.
Maxims thus either pass the test, or fail in manner (a) or (b). To make the connection between the result of the test and duty, one might proceed in one of two ways. One might say (i) if M passes the test, one has a duty to act on M, or (ii) if M fails the test, one has a duty not to act on M. These are not, of course, equivalent, and it is obvious by inspection that (in this mode of application) (ii) is the correct version. (i) would allow the possibility of contradictory duties: when A and A* would be equally good ways of getting E in C, it would be obligatory to act on both the corresponding, incompatible, maxims. Besides, Kant says that (ii) is what he intends:
If the intent of the action can without self-contradiction be universalized, it is morally possible; if it cannot be so universalized without contradicting itself, it is morally impossible.(9)
Clearly, then, it is the failure of a maxim that matters for Kant. It is permissible to act on a maxim that passes, but no duty attaches to it. The other moral notions on which Kant lays great stress should presumably take a similar reading. Having a good will would consist in being prepared not to act on maxims which fail the test; acting out of duty would be taking the failure of a maxim as a conclusive reason for not acting on it (other reasons might, of course, be concurrent).
The ways in which maxims may fail the test serve to draw, for Kant, the distinction between perfect and imperfect duties, one dimension of the four-way division of duties in the Grundlegung (the other is the distinction between duties to oneself and duties to others). A perfect duty arises in connection with a maxim which could not conceivably be a universal law, an imperfect duty with a maxim which could be a universal law, but which would not be willed to be such by a rational agent. The division between these two types of duties - at least in the case of duties to others - is not without importance. Perfect duties to others involve rights, allow no exceptions in the interest of inclination, and may be enforced by a penal authority: without stretching too much, we could say that the domain of perfect duties to others is the domain of justice. The distinction between perfect and imperfect duties to others - and in this paper I will leave the thickets of self-regarding duties unbroached - is, I take it, intuitively recognized: to provide a sound theoretical underpinning for it, even if he only did that, would be a major achievement for Kant (and would perhaps raise the level of discussions of "the enforcement of morality").
To draw the distinction, and to understand the Categorical Imperative, we need at least to get clear about two things:
1. How, exactly, is the contradiction supposed to arise in the supposition that a given maxim be a universal law?
2. How is it possible to warrant necessary truths about what rational agents would will to be universal laws?
(1) is the problem that I will, herein, be concerned with most. Perfect duties to others seem the most important of the four types; and, I hope, an answer to (1) will break the ground for the proper treatment of (2). Many of the pitfalls in the way of (2) are the same as those in the way of (1).
Kant's example of a perfect duty to others, in the Grundlegung, is the duty not to make false promises. The case is familiar: a man is in need of money, and can obtain it by borrowing if he promises to repay the debt by a fixed time. He knows that he will not be able to return the money, but promises anyway. According to Kant, his maxim would be: "Whenever I believe myself short of money, I will borrow money and promise to pay it back, though I know that this will never be done". Presumably the end to be served is security, welfare, or whatever. Now how does the transmogrification of this maxim into a universal law bring with it a contradiction? Kant says:
I then see straight away that this maxim can never rank as a universal law of nature and be self-consistent, but must necessarily contradict itself. For the universality of a law that everyone believing himself to be in need can make any promise he pleases with the intention not to keep it would make promising, and the very purpose of promising. itself impossible, since no one would believe he was being promised anything, but would laugh at utterances of this kind as empty shams.(10)
On first reading, Kant seems to be suggesting that the reason M could not be a universal law is that if it were, there would be no promises. Generalizing, it would seem that he intends that a maxim "For E, I'll do A in C" cannot be a universal law if universal adoption of the maxim would have the consequence that there would be no A's. But that cannot be what he means: it would be too arbitrary. If someone said, intending it as the single unelliptical criticism of our conduct, "If everyone adopted that policy, there wouldn't be any A's", that would hardly be calculated to turn our bowels liquid with guilt. A putative fundamental moral principle must make coherent sense to those for whose judgements it is supposed to account. The principle that one ought not to act on maxims whose universalization would entail the nonexistence of the actions mentioned therein, taken by itself, does not make that kind of sense.
I think it is clear enough what Kant has in mind. He often inveighs against the immoral agent as one who makes exceptions in his own favour. This charge seems one that has at its root the idea of fairness. The man who acts on a maxim that is not fit to be a universal law is skimming off an advantage he could not will to be available for everyone. Now wherein lies the advantage? What does a man gain by using a maxim unfit to be law?
Clearly he gains what acting on maxims gains one! The content of a maxim is a rational strategy for achieving an end; the advantage one gets from using a maxim that could not be a universal law is the end that that maxim is designed to serve. The force of saying that the advantage is unfair is that that strategy for gaining the end is one that could not be followed by everyone. The strategy works only because generally not followed. The interpretation of "self-contradiction", so far as the application of the Imperative is concerned, should, I suggest, be as follows: A maxim is self-contradictory if universalized if the supposition that that maxim be a rational strategy for obtaining the end mentioned in the maxim, for everyone, is self-contradictory. The supposition will be self-contradictory if it follows from it that not everyone can gain the end in question by acting upon the maxim (assuming the causal beliefs which led to the adoption of the single-person maxim are true). This, I think, captures the sense in which someone violating the Categorical Imperative is being unfair in making an exception for himself. The rationality of the strategy he adopts depends on its being a strategy not adopted by everyone; what makes it a good strategy is that it is idiosyncratic.
Taking the contradiction to arise only if universalizing a maxim would result in there being no A's would lead to anomalous results, if one viewed the Categorical Imperative as a kind of analysis of fairness.
On the "no A's" reading, a maxim would pass the test even if some agents could not obtain E if M were universal law, so long as some A's were still possible. This, of course, simply points up the arbitrariness of the "no A's" reading: why should the preservation of the possibility of performance of A's (even perhaps, by only one person) make the difference between a passing and failing maxim? The reason why a maxim should fail if at least one person could not gain E by it is, on the other hand, reasonably clear. Consider the following maxim:
Mb: For the end of painless release of hostility, when I find someone weaker than I, I'll beat him up.
I do not mean to suggest that this is a maxim a rational agent could plausibly be supposed to adopt. Indeed - and I repeat that so far we are dealing with a first approximation to the application of the Imperative - not even the examples Kant gives will turn out to be such. Suspending these scruples for the moment, we may ask whether Kant would intend a maxim like Mb to pass the test. Obviously, he would not, yet the "no A's" interpretation would take him so to intend. For if Mb were a universal law there would be beatings up enough to outdo a disorganized chicken-yard. But at the end of the line would be poor hapless Chicken Omega. Her fate is enough, on my reading, to flunk Mb. And surely Kant would concur.
Cases where asymmetric transitive relations are essentially involved in a maxim submit easily enough to proof that if everyone adopted the maxim, it would fail to be a rational strategy for all. Other, less artificial, cases are less likely to go so easily. Kant's own example, while convincing, is not easy to crack. To it I now return.
In Kant's treatment of the maxim of false promising he does appeal to the fact that there would be no promises if the maxim were universal law. That is not inconsistent with the present interpretation, of course: if there would be no promises, no one, and thus not everyone, could gain an end by making a false promise. But it is not entirely clear why Kant thinks there would be no promises. I will not attempt to construct an argument proving that there would be no promises, though I will make an occasional suggestion. I suspect that a water-tight argument would require a number of lemmas whose discovery and justification would be difficult and time-consuming, even if, ultimately, illuminating. I will, however, make some comment on possible interpretive pitfalls.
An interesting argument by C.D. MacNiven(11) would locate the impossibility of promises under the universalization assumption, and more generally the impossibility of a given type of action whose maxim is associated with a perfect duty, in the conventional nature of such actions. That is to say, there are certain actions, of which making promises can serve as a paradigm, which are possible only because there exists a shared practice or institution among a group of people. The practice could be viewed as constituted by a system of accepted rules; without the acceptance of the rules there would be no practice, and, pro tanto, no promises. For a community to have a practice, or institution, of promising, is for it to have some form of "commitment behaviour" or a signal, which, when attached to the description of an action, indicates, at the very minimum, that the probability of the performance of an action answering that description is increased over what it would be had the signal not been made. In a way, then, for a signal to have the sense of "I promise ..." it must be governed by a rule tantamount to "Keep promises, ceteris paribus"; an expression used in a particular group of people, if not so governed, is not the local equivalent of "I promise ...". A person who adopts the "false promise" maxim is rejecting the rule which requires promises to be kept. If everyone adopted the maxim the rule would cease to be accepted at all, and there would then be no promises. False promises are thus parasitic upon a general practice of keeping promises. On this line of interpretation, perfect duties would arise when a maxim to be tested ran counter to a constitutive rule for the action mentioned in the maxim. The distinction between perfect and imperfect duties would be based on the distinction between action-descriptions which do, and those which do not, rest on a background of accepted practices.
Now for all that the distinction between nature and convention is the pons asinorum of ethical theory, this reading of Kant on perfect duties cannot, I think, be right. It would entail that one has a perfect duty to abide by whatever conventions one encounters, and one surely has no such duty. If the Categorical Imperative is anywhere near correct it must distinguish between a promiser's "conventional obligation" to keep his promises and a slave's "conventional obligation" to obey his master. I prefer for the moment to operate on the heuristic assumption that Kant's theory can be shown to be correct, so I will take it that the fact that the MacNiven interpretation has this unacceptable consequence is decisive against it.(12)
Another reason for suspicion about the "practice" interpretation of perfect duties is that it has not been conclusively demonstrated that promising does presuppose an accepted conventional practice. One might attempt to analyze the concept of a promise without invoking such conventions. A first stab might run as follows:
x promises y to do A = (a) x does B intending to produce in y the belief that x intends to do A unless Z, (b) x intends the belief to be produced partly by recognition of the intention in (a), and (c) x intends to produce the belief in y that the intention to do A unless Z is conditional on the success of intentions (a) and (b).(13)
I have no doubt that this is not a correct analysis of promising, but something like it may come close. The role of conventions in promising, on such an analysis, would be to facilitate the production of the beliefs in question. But promising would be a natural act, getting someone to believe that what you intend to do depends on what you get him to believe you intend to do, something that would be possible without conventions (if the participants were bright enough).(14) A false promise maxim would treat the success of intentions (a) and (b) as irrelevant to the intention to do A unless Z; thus if all rational agents knew that a false promise maxim were universalized, no one could succeed with intention (c); and if they all knew that, they couldn't even try, so they couldn't promise. Deriving the contradiction in this way seems to require universal knowledge that the maxim is universally adopted; I do not know whether this is part of Kant's intent.
So far, then, it is not clear how Kant's false promise example is to be taken. The way in which I find it more reasonable to take it, in which promising is viewed as a natural act which may be facilitated, not constituted, by conventions, will treat promising as a species of some genus of interaction between rational agents, and will so treat the duty on question. The duty not to make false promises may be an instance of some more general duty having to do with co-ordination of activities.
A second kind of problem area is precisely analogous to one which proves so vexatious for those who put forward generalization principles. The problems derive fromthe fundamental problem of admissible descriptions. Consider the following maxim:
Md: For happiness and security, I will become a doctor and practice medicine.
If this were a universal law, someone would fail to get happiness and security by practicing medicine, likely everyone. Yet one hardly wants to say that one's doctor is immoral for adopting early in life the policy of pursuing his happiness by practicing medicine. Especially would one be reluctant to say this since it applies in spades to entering the groves of academe.
Of course we need not suppose that our doctor acts on Md; he may be acting on:
Mt: For happiness and security, I will find an occupation for which my talents are suited, and work in it.
We could say that doctors who act on Md are evil, while those who act on Mt are good; but we had better have a reason for so saying, other than the fact that our test is ruined if someone acting on Md is not immoral.
Another type of problem is posed by Hares's fanatic. What does our test say about the Nazi's maxim:
Mn: For the elimination of Jewry, when I find a Jew, I'll kill him.
If Mn were a universal law, no one who desired this end would necessarily fail to achieve it. What better way of eliminating Jewry than having everyone adopt Mn? Yet actions flowing from such a maxim are clearly immoral.
The road through these problems begins with a reconsideration of what maxims are supposed to be: they are rational policies for achieving ends. It is that, I think, which will provide something analogous to criteria for admissibility for descriptions of actions. Consider again Md and Mt; clearly Mt is, in some sense, a more sensible policy for a person to adopt: someone acting upon Md might be thought fanatical, or at least oddly single-minded. So too would someone be thought who adopted the policy of always making false promises when financially embarrassed. If there are easier ways of picking up cash, wouldn't it be better to make use of them? If the prospective MD encounters a snag in medical school, should he simply press on? What if he discovers an unsuspected talent for law or engineering, or a hidden facility at the production of smarmy verse?
The point of these questions is to bring out that the maxims we have so far looked at are either foolishly irrational if looked at as complete maxims, or very minor detailed excerpts from reasonable maxims. If taken as someone's policies for gaining his ends, they are fanatical, and stupidly so.
The policies (policy?) of a rational agent would not resemble these maxims. What they would look like can be teased out by noting that in attributing an intention, or a policy, to someone we are saying something not just about what he does, but about what he would do if certain contrary-to-fact conditions obtained. Suppose a person does A; he does it, say, to get E. He would not have done A under a variety of conditions: if he had not wanted E, if doing A would thwart a more important end than E, if the circumstances were such that (he believed that) doing A wouldn't get E ... .(15) All this is contained in the judgement that someone did A in order to get E. A maxim, as the formulation of one's policy concerning what one will do in order to get certain ends, must encompass a range of counterfactuals at least as broad. Explicitly or implicitly, the maxim "For the sake of E, I'll do A in circumstances C" must be understood as surrounded by this large set ofqualifications. Now one could, when outlining the total content of a given maxim, attempt to enter these qualifications in detail; one could say "unless I cease to want E, unless doing A interferes with other ends of mine, unless doing A would not be a way of getting ...". Every other maxim would have to be treated in the same way, of course, since the various ends of an agent, and the possible means to obtaining them, would form an interlocked system. The final result of this process of qualification would be, if it could be stated at all, an extremely complicated single maxim, with subclauses extending over the horizon.
One could leave matters in that state, saying that the single maxim of a rational agent would be very long and complicated, and impracticable to state. But what would be unsatisfactory. A list of qualifications would not wear its heart on its sleeve: it would not make clear its own rationale. Why these qualifications rather than others? The answer is, of course, obvious: a rational agent chooses paths of action which will, so far as he can ascertain, gain for him as many of his most important ends as possible, while avoiding as many as possible of the things he most wants to avoid - in brief, actions which will maximize utility. A test for the completeness of the complex single maxim is whether it corresponds with a general strategy for all possible combinations of ends and causal situations, for maximizing outcomes. But even if complete the single maxim merely characterizes the strategy, as it were, extensionally; it does not give the sense of the required policy. That is given by:(16)
Ma: I will act so as to achieve maximum consistent harmony of my ends.
If put in imperative form, this is what Kant calls the assertoric imperative, theimperative of happiness. Maxims of the kind Kant serves up as examples, maxims corresponding to hypothetical imperatives, must be seen as qualified and constrained by this maxim of happiness. Specific ephemeral intentions, and indeed actions, could be seen as flowing from the maxim of happiness in the manner of Aristotle's practical syllogism.
The policy would be that given by Ma; actions and intentions would be the realizations of that policy in particular circumstances by agents with particular ends.
In a sense, the demand the leads us to Ma, the maxim of rational agents, is the demand that a rational maxim be a "life-policy" satisfying the condition that following it would be a good way of achieving one's ends and not just accidentally a good way. The fact that the policy is an agent's intention with respect to his life requires that it be one which would be sound even were certain empirical conditions different, that is, it must have "counterfactual content". To be sure, it need not encompass all possibilities: it need not yield a consequent for every logically possible antecedent. It need not generate an intention for the condition "If I were not a rational agent ...", for example. Perhaps we could say, preserving the vagueness appropriate to my ability to say anything right about this question, that the policy should yield intentions for all possible conditions under which the agent "would still be the person he is". This is a fairly liberal requirement; it does not rule out "If I were a dolphin ...", but does (so far as I know) "If I were a delphinium ...".
Even if the demand for a !life-policy" does not force us to something as simple and powerful as Ma, it does require that the policy arrived at satisfy two conditions:
1. It will make no reference to the determinate content of one's ends; it willbe designed to obtain one's ends whatever they turn out to be. This seems to me to be in the spirit of Kant: from the point of view of Reason ends, are not worthwhile because of their content, but because they are ends some person has. (This is why rational agents are ends in themselves: they are the unconditional source of the conditional value of the objects of their desire).
2. A high level of generality is involved. An overly specific maxim could not encapsulate the counterfactual content necessary to meet the rationality requirement. A maxim which, say, made reference to specific characteristics of the agent would be irrational (suppose one adopted the maxim "For my happiness, I will found a state dedicated to promotion of the ends of Aryans"). That is not to say, however, that a maxim which included, or entailed, "If I have a property F, which few others have, I'll get everyone to promote the ends of those who have F" would be irrational.
What provides criteria for admissible descriptions, or the analogue thereto, is the requirement of rationality of policy, along with the intentionality of intentions, policies, and the like. The admissible description of an act would be the most general of those under which an ideally rational agent would be performing that act; it would of course have to describe the act as directed toward gaining an end. The general description would, so to speak, give the sense of the network of ceteris paribus clauses which any rational agent would tacitly acknowledge to be contained in the intention from which the action flowed.
The criteria for admissible descriptions - or better in this context, simply the rationality requirement - must be satisfied before a maxim is brought to the Categorical Imperative test. Thus some maxims which allegedly might create problems for test may not even be candidates for testing. I will, of course, be discussing the way in which the test is to be applied to rational maxims; but it may be worth noting here some effects of the rationality requirement. Some difficulties are avoided by making it explicit.
For example, MacNiven's interpretation would be untenable. Just as there would be no reference to the determinate content of one's ends in rational maxims, there would be no reference to the determinate content of the conventional practices which, as a matter of fact, surround a particular rational agent. Something more general, such as "When there are conventions of type X ..." might be included. (Even this might be thought too specific, but I will be taking up that question shortly). Further, a policy of whole-hearted allegiance to whatever conventional practices one encountered would hardly be rational. That this is so creates a neat double bind for a slave owner. Suppose for the sake of illustration that a slave owner's maxim must be one of the following:
(a) When there is a property F, which I have but few others have, I will get others to work for the satisfaction of the ends of those who have F.
(b) When the conventional practices of my community assign me a role with benefits and burdens attached, I will accept those benefits and burdens, and require others to do so for their roles.
(a) would be reasonable enough, but would (I assume) fail the Categorical Imperative test. (b) would fail the rationality requirement. A policy of accepting whatever role one finds oneself in is far from optimal - it entails being a willing slave or beast of burden if that's how things work out. A slave-owner who swears that he would have been a well-behaved slave had fortune so decreed may get marks for sincerity, but is too naive to be taken seriously. Sincerity is the cheapest of the virtues, and a sincerity that naive is almost enough to forfeit one's status as a rational agent. In brief: it's stupid to act on a policy which would permit your being exploited; it's immoral to act on one which allows you to exploit. So you can't defend yourself from the charge that you're immoral when you exploit by saying you'd have played the game had roles been reversed.
Now to return to the way in which the Categorical Imperative is to be applied to rational maxims. Clearly a maxim which passes the initial rationality-generality test is intended to be modified by the universality test, or, perhaps better, a maxim which is to be acceptable must pass the universalizability test absolutely, and pass the rationality test to the degree that so passing is consistent with universalizability.(17) Morally acceptable maxims must, so to speak, compromise their effectiveness as rational strategies in order to assure that they may - in principle - be employed by everyone. For those possible actions where no other rational agents are involved, Ma will clearly satisfy both the rationality and universalizability requirements. Indeed, one might try to make a case for saying that Ma would unqualifiedly pass the Categorical Imperative test: it does not seem to be necessarily the case that if everyone acted on Ma someone would fail to achieve his ends. But to argue thus may be to miss the point. The test, in essence, introduces a distinction between two categories of end-pursuing actions, those in which one acts solely upon inanimate things, and those in which one acts upon, or through, other rational agents. In the "counterfactual content" of Ma would be a number of subordinate strategies dealing with cases where one might achieve one's ends by acting upon another rational agent; some of these strategies would be, intuitively speaking, exploitative. Submission of one's maxim to the Categorical Imperative should eliminate these exploitative strategies. One is obliged to construct a maxim to cover the cases of interaction with other rational agents whose content does not include such strategies.
In effect, then, the drive toward generality required by the condition that one's maxim be rational is reversed, for one category of actions, by the condition that one's maxim pass through the filter of the Categorical Imperative.
The selection for special treatment of those actions where one interacts with others requires that some of the specific content of Ma be brought to the surface for possible modification. One must extract from one's maxim those strategies which have to do with what one will do with and to other rational agents. It is these that play a central role in applying the test, for only among these (assuming the rationality requirement already to be satisfied) could any contradiction arise in the supposition that everyone gain his ends by following them.
The question is, of course, how specifically must maxims be rendered for the purposes of the test? The thrust of these reflections so far has been in favour of maximum generality; presumably any relaxation of generality must be the least possible, and must be justified completely by the considerations which led away from the level of generality imposed by the rationality requirement. Now what introduces the need for greater specificity is simply the acknowledgement that interactions with other rational agents are to be given a special status. So, it would seem, the degree of specificity introduced will be only that necessary to separate off a strategy for interacting with rational agents from a strategy for dealing with inanimate things. The descriptions involved will make reference to rational agents qua rational agents; since they need be no more specific than that, they may be no more specific.
The types of interaction which may be cited in a maxim brought to the test will be only those which can be characterized by reference to the bare concept of rational agency. That is, the admissible descriptions of forms of interaction will be descriptions applicable to any sort of rational agent: they will not be drawn from a stock applicable only to agents of some given empirical character. The relaxation of generality proceeds only far enough to take account of that status in virtue of which a creature is an object of moral concern. The status is that of a conscious person, desiring ends, capable of deliberation and planning about the means to pursue those ends.
The effect of this modification of Ma should be to produce a compact set of strategies governing one's interaction with other rational agents. The requirement of rationality still stands, of course: these strategies must be rational ways of obtaining one's ends, and they must have much "counterfactual content": they must be reasonable even under a variety of contrary-to-fact conditions. Thus reference to special features of one's nature and condition are debarred. One's plans for interaction with other agents must satisfy the condition that they be good even were one a drastically different person.
The policy which a rational agent would adopt under these strictures would be very like that which would be chosen by a rational agent under conditions of ignorance. One could derive the maxims which would satisfy the twin conditions of rationality and universalizability by postulating a group of rational agents ignorant of any facts about themselves except that they are rational agents, charged with the creation of a "life-policy" to which everyone must agree, and which will be binding on all once agreed to. Looked at it in this way, Kant's moral theory is a generalized version of Rawls' work on justice.(18) Rawls imagines a group of rational persons, ignorant of their roles in the institutions of their society, gathered together to decide upon principles for the criticism of those institutions; thereafter all complaints against the institutions will be judged according to the chosen principles. Kant does not present his theory in this way, though the formula of the Kingdom of Ends has a Rawlsian flavour. But the two accounts are obviously closely related. One could set out the procedure for generating the maxims that pass the Categorical Imperative by constructing the policies which would be adopted by rational agents under the following conditions:
1. They do not know what they will want, just that they will want.
2. They do not know what empirical realization they will have.
3. They know that they may be able to interact fruitfully (this is presumably involved in their knowing that they are rational agents).
4. No life-policy will be embarked on until all rational agents agree to one which thenceforth is binding.
What, after all, could be taken into account when deciding, under these conditions, whether a proposed policy should be accepted or rejected? Simply, I should think, whether it is a policy everyone could use in pursuit of his ends. In conditions of ignorance, the Categorical Imperative test would be the only filter rational agents could use to strain out policies not acceptable to everyone. An extension of Rawls' account of justice would seem to promise results equivalent to Kant's theory of morality.
But the Kantian way of reaching those results is the correct one. Both viewpoints see morality as rational self-interest constrained by impartiality, but the development of Rawls' theorem leading to the two principles of justice - and the more general form alluded to in the preceding - require the introduction of postulates of ignorance (about one's social class, talent, and so on) in what could be thought an ad hoc way. Of course the postulates are not wholly ad hoc; Rawls' rational agents in the original position are conceived to be ignorant of those features of themselves which are deemed to be irrelevant when questions of justice arise. Proceeding in this way, however, is less than optimal - the account of justice is made to seem dependent on an external criterion of relevance. Ideally, the analysis of morality should generate the criterion of relevance. And that is what the Kantian analysis does. What would be derived in the Rawlsian account from postulates of ignorance flows naturally in the Kantian from the requirement of rationality.
In the Kantian account, an agent is not assumed to be in ignorance of his position and talents, but the requirement that the policies have counterfactual content, that his actions on a particular occasion follow from a policy which would be rational under all possible circumstances, entails that his maxims must be stated as if they had been formulated in conditions of ignorance. The Rawlsian account may bring the Categorical Imperative "closer to intuition", in Kant's phrase, but the purely Kantian one must at least be seen behind the scenes justifying the restrictions placed upon the deliberations of the agents choosing life-policies.
When making a moral judgement about a particular occasion, then, one must decide, on the basis of one's knowledge of the agent's desires, beliefs, and so on, what very general policy a fully rational agent would be acting upon were he to act thus with such beliefs and desires. That policy is the one the agent in question must, qua rational, be deemed to be following. If it does not pass the test the action is wrong, whatever the agent - sincerely or insincerely - represents his maxim to be. This does not conflict with what I stipulated earlier, that the maxim to be tested must be the maxim the agent actually acts upon; any rational agent can see, on reflection, what the rational policy underlying an action of his would be - if he maintains that really he is following a different, irrational policy, I do not see that we cannot simply count him as in bad faith. If he is incapable of seeing what the rational policy would be, it is not clear that he is rational.
To close, I will attempt to sketch the way in which maxims involving interaction between rational agents may be adjusted so as to pass the test. This is not entirely a simple matter, since it is difficult to meet the standards of generality necessary in the formulation of maxims. Presumably the manner in which the types of interaction between agents ,may be specified is to be restricted to a characterization of the effects their actions may have on the capacity of each other to obtain their ends - in, of course, the most general terms possible. Indeed, one could argue that the passing maxim having to do with interaction should simply be "I will never treat another solely as means, but always as an end". That would be general enough to satisfy anyone - but it cries out for a codicil explaining what it entails. I will assume that, so long as the characterization of a maxim is general enough to apply to all possible rational agents, I may subdivide forms of interaction, and assign a clause to each. Or, alternatively, what I will present may be thought of as a kind of codicil to the formula of the end in itself.
For present purposes, I will adopt a fourfold classification of ways of interacting, two "negative" and two "positive". The classification is probably not in any way fundamental; it simply is a natural one in this discussion. Neither is it an exhaustive division of relevant ways of interacting: I omit, for example, any consideration of the vexatious question of actions restricting another's liberty of thought or speech. The four types I will examine are: (1) interference, (2) competition, (3) co-operation, and (4) gratuitous assistance. These will be taken up in turn.
1. Interference. I will understand by this, gaining one's ends by partially or wholly, temporarily or permanently, restricting another's ability to pursue his ends. This characterization may be broad enough to subsume competition, but I treat that case separately, since it raises special problems. Paradigm cases, for human beings, would be killing another, injuring him, or curtailing his freedom. (I should say again that I am dealing only with cases in which the maxims of the person with whom one is interacting are themselves all right). Interference seems necessarily to involve an asymmetry of agent and patient. If a maxim permitting interference is acted upon, someone loses because of that action; the patient is unable to gain what he could have gained had there been no interference. In the cases an interference maxim would be designed to cover, it would be a maxim strategically rational for only one party. Prima facie, Ma as modified by the Categorical Imperative would contain a clause tantamount to "... unless I would be interfering with another". Of course, this does not appeal in cases where interference is the only policy either agent can follow. Kant discusses, in The Metaphysics of Morals, "The Right of Necessity", the example of two shipwrecked men struggling for control of a plank which can bear only one of them to safety. One or both of the men must die, and their positions are wholly symmetrical; is it permissible for one to push the other away from the plank? I think we must simply say for such cases that "permissible" is undefined; applying moral principles to such cases is, if you like, like dividing by zero.
It might be tempting to modify the clause further, along utilitarian lines. Cases in which interference would be a rational strategy for one person could be divided into two main categories: (a) those in which the loss to the patient is greater than the gain to the agent, and (b) those in which the loss to the patient is smaller than the gain to the agent. A plausible case could be made for permitting interference in cases of kind (b), but not of kind (a). Modifying the maxim in this way would generate something which would optimize the combined utilities of the parties in a case where interference would be a rational policy.
Such a maxim would constitute a policy which everyone could use to gain his ends, in such a case, so far as the natural circumstances permitted ends to be gained.
But this line of argument would be mistaken, I think. Strictly speaking, it is not true that both parties in an "interference" case could use the "utilitarian" strategy. By the nature of the case, only one could. That is, the natural circumstances are such that someone must take a loss. Even for someone operating on Ma unqualified there are possible circumstances in which his ends, or a number of the most important ones, could not be achieved. That kind of impossibility is not caused by a maxim a person has; it is natural. The impossibility which is relevant to the Categorical Imperative is an impossibility which results from the maxims supposed to be adopted by everyone. The asymmetry produced by a maxim permitting interference does introduce this kind of impossibility: the patient cannot use the maxim. The losses suffered by the person debarred from using a maxim permitting interference are due to natural circumstances. He cannot , so to speak, pass on to someone else the hindrances nature has put in his path. The hindrance belongs to him who encounters it. That is not to say, of course, that the Imperative will not require that persons co-operate in sharing the burden of natural hindrances, nor that it will not require assistance to be given; it is simply to say that to interfere with someone solely on the ground that to do so is prudent is an unacceptable action. It might be legitimate to interfere with someone in circumstances of the kind we have been discussing if such interference were justified by the fact that he ought voluntarily to have promoted one's ends. But that belongs to the next round of epicycles.(19)
2. Competition. This type of interaction bears certain analogies to interference. It involves cases where two or more agents desire the same thing, or desire portions of a scarce good. In particular instances of competition, I assume, not everyone can get what he wants, or in the simplest cases, where two persons each want E, only one can have E. Unlike interference, competition is symmetrical; if x is competing with y for E, y, willy-nilly, is competing with x. So one cannot simply qualify one's maxim with "... unless I would be competing". If this were universal law, neither competitor would gain his ends in competitive situations.
A passing policy to deal with competition must, I think, be a policy for the long term, not specifically for individual instances. In the individual instances someone has to lose. And of course Ma itself is a long term policy, designed to gain one happiness; individual cases of competition would be for parts of, or means to, happiness.
Now the ideally rational policy to govern competition over the long term would be to devise a set of conventions which would regulate the behaviour of those subject to them when engaged in competition. These conventions would, of course, be subject to moral assessment; but the criteria for such assessment would be - so far as I can see - the same as those for conventions governing distribution of benefits in general. They would be assessed in the same way that co-operative practices would be assessed. Indeed, competition regulated by conventions is a species of co-operation. So the maxims governing conventionally regulated competition can be dealt with under the heading of co-operation.
Natural competition, I will say, is competition without such regulating conventions. Rational agents may find themselves involved in competition under circumstances where "rules of the game" cannot be, or have not yet been, established. A clause dealing with such circumstances must be written in to the modified maxim. The clause will, no doubt, mirror some features of the criteria for the assessment of conventional practices. For the special case of competition the policy should at least satisfy the following criteria:
(a) Interference is not involved.
(b) The policy should not make the cost of getting the scarce good excessive.
(c) It should bring about a stable outcome in particular cases. (This may covered by (b)).
(d) It should guarantee against perpetual losers. That is, even though a winner and a loser are unavoidable in particular cases, the long-range policy should assure that asymmetries are not permanent.
These requirements in effect stipulate that the policy should be nonviolent and stable, and should, over the long term, assure access for all to a share of scarce goods. A policy which may meet the criteria is:
"... when there is natural competition for scarce goods, I will compete, but will be content with a sufficiency; and I will not take from another what he has already got unless it exceeds a sufficiency."(20)
The second half of this statement may belong, properly, to the policies concerned with the treatment of those who act on unacceptable maxims, in this case those who take more than they ought of a scarce good. The policy is an open-textured one: what a "sufficiency" amounts to will have to be determined by the facts in particular cases. Relevant features of the case would be the scarcity of the good, its necessity, and so on. Obviously, this is as it should be: a moral theory which entailed that correct behaviour in such situations was not a complex function of the facts would be an inadequate moral theory.
3. Co-operation. Co-operation includes a range of cases, from those in which one person acts to further the ends of another, on the condition that the other act to further his, to those where a large number of people jointly act to produce some good, which may be distributed among them. Co-operation is fruitful if the participants will better serve their ends by joint action than they would were they to act independently. Obviously, it is rational to participate in co-operative activities when they are fruitful. The problem is: what policy with respect to the conditions of one's participation should one follow? Rational self-interest would presumably counsel holding out for as large a share as possible of the fruits of co-operation, being prepared as a last resort to settle for something less. But this strategy of attempted blackmail could not be successful for everyone. A strategy of taking whatever is assigned by way of benefits to one's role in a co-operative structure might be universalizable but would not be rational. The twin demands of rationality and universalizability seem to drive one in the direction of equal shares for full participants, much in the way that one is driven in Rawls' account of justice. Equal shares, as a first approximation, would be modified in the sense of Rawls' Difference Principle. Thus the passing maxim would seem to be something like:
When co-operation is fruitful, I will do what is required of me by the structures of co-operation, in return for benefits, so long as benefits are distributed in proportion to the degree of participation, unless a departure from proportionality will work, for reasons not involving bad maxims, to the advantage of all participants.
Degree of participation is a notion not yet defined. It is included not to encourage tail-chasing debate about whether garbage collectors contribute more to society than scandal-sheet journalists, but to point to the possibility of diverse degrees of involvement in a particular structure of co-operation, especially where the structure is temporary, and where participation is entirely voluntary. For example, suppose two men co-operate in the removal of rocks from a field. One spends three weeks removing all the rocks that can be taken by one man alone; the other joins him for three days to remove those that require efforts of two men. Even assuming the field to be worthless unless completely cleared, one man has participated more then the other (and intuitively deserves a greater share of the cleared land).
The qualification that the greater efficiency attendant on unequal shares not be a result of the adoption of bad maxims by some participants again points beyond, to the maxims to be adopted regarding the treatment of those who act immorally.
But it needs to be incorporated here, since the Difference Principle would not be acknowledged for those cases where blackmail was at work. A self-interested rational agent (or a group of agents) could arrange matters so that co-operative practice would be made more efficient if he were granted greater shares of the fruits of co-operation; he could deliberately go slow under a régime of equal shares. The question whether possible greater efficiencies would be due to deliberate blackmail or to innocent causes is crucial to the question whether inequalities would be acceptable.(21)
At this level of generality, nothing need be said in the maxim about possible reasons why inequalities might benefit everyone. Those reasons belong in the empirical part of moral philosophy, after a group of rational agents, so to speak, has discovered who they are, what their empirical realizations are. The Difference Principle, at this level, is included merely as an abstract possibility.
4. Gratuitous assistance. Here not much needs to be added to Kant's own discussion, except to correct an unsympathetic misinterpretation, and to point to some difficulties. The misinterpretation is illustrated by Frankena:
It is true that if one adopts the maxim of not helping others in need and wills this to be a universal law, he is likely to find himself willing inconsistently to abrogate this rule, since he is likely himself to be in need sometime. Still, it is not hard to imagine a man whose fortune is fairly sure or one who is willing to take the consequences of his maxim's being universally acted on; if there are such people ....(22)
Obviously neither of Frankena's cases will jointly satisfy the rationality and universalizability requirements. A reasonable policy might be: If able to give help, I will not give it; if in need of help, I will accept it. But such a policy could not be a way of gaining ends for everyone, since the first clause guarantees the uselessness of the second. A maxim "I will never give help, nor accept it" would be a fool's maxim. Of course there are fools; but if I have said anything here, it is that Kant is not interested in what fools do will, he is interested in what rational agents would will.
The Kantian question is "What policy for every possible circumstance would it be rational to will?". Such a policy has to be more general than a specific policy adequate for one's given empirical circumstances, and it is as a general policy for all possible circumstances that one has to see it when deciding whether to will it to be a policy for all rational agents. Kant may have contributed to the chance of misinterpretation here, since he suggests in his discussion that it would be psychologically impossible for someone to will that no-one ever help him. But I think this is a mistaken reading of Kant's discussion of the case. He is concerned with what it would be rational to will to be a universal law, not with what an only partly rational human being could, as a matter of psychological fact, will. It is possible for a human being to be in bad faith, to refuse to recognize the irrationality of what he wills, and thereby to think that he satisfies the requirements of the Categorical Imperative.
I believe that Kant would hold the first three of my categories of interaction to give rise to perfect duties to others, the last to an imperfect duty. I have some sympathy for that view, but I am not sure that it can be made out satisfactorily in the way Kant would try to make it out. The difference between a perfect duty and an imperfect duty is supposed to arise out of the difference between maxims which could not be successful strategies for everyone, and those which could, but which it would be irrational to will to be everyone's strategy. But in the case of a maxim of unhelpfulness, arguments similar to those given in the cases of interference, competition and co-operation can be given. Astrategy of neither giving help nor taking it would not be a way of achieving ends for someone in need of help. Of course, it is not necessarily true that someone will need help; but neither is it that someone will be in a position to gain from interference. But if someone is ever in a position to gain by withholding needed help, someone will lose if the help is withheld; there is an asymmetry nevertheless. This is a question needing further work. I suspect the answer lies in Kant's view that a violation of a perfect duty is a hindrance to another's freedom to pursue his ends, but of an imperfect duty, merely a failure to promote his ends for him. Whether something along these lines will march is a matter I cannot settle here.
The obscurity of the distinction between perfect and imperfect duties is deepened by the fact that the boundaries between the categories of interaction are not sharp. In particular, this is true of the boundary between co-operation and help. The blurring of the line between the two is due, I think, in great part to the fact that the line is drawn by conventions. Whether one person's aiding another on a given occasion counts as acting according to the understandings of a co-operative practice, or as gratuitous assistance, is a function of what the co-operative practices are of the group to which the two persons belong. These understandings may be extremely vague and open-textured, so that it may be next to impossible to tell whether an act open to one is gratuitously helpful, or obligatory under a just set of conventional understandings.(23) (Or perhaps one should say that the action simply would not be a clear case of one or the other). I suspect that our own conventions are none too clear on this matter: we take it that one has a special duty to help those who have helped us; this would not be reasonable either on the assumption that help were gratuitous or on the assumption that help were required by fair conventions - it would be absurd to hold that one had a special duty to repay debts to those fortunate enough to have been in debt to one in the past. But if another's act of assistance were seen as bringing local clarity into one's relations to him the special duty might make sense.
This brief discussion of "interaction maxims" is, of course, not complete; the arguments that only certain maxims would pass the Categorical Imperative test are not conclusive. But I hope that the case has been made that the application of the Imperative must proceed along the lines of those arguments. The case rests on the more fundamental one, that maxims must be treated as rational strategies for the pursuit of one' s ends, that taking seriously the requirement that they be rational strategies forces them to be stated in a very general way, and that this generality in the statement of maxims is what provides the analogue in the case of the Categorical Imperative for the criteria of admissible descriptions in the case of generalization principles. If the benchmarks I have been trying to establish are placed rightly, a Kantian map of the moral terrain is not in principle a difficult task.
I hope too that the way of looking at Kant that I have been advocating helps to bring together Kant's ethical theory and Rawls' theory of justice. Whether Kant's theory as presented here is corroborated hypothetico-deductively by taking Rawls' as a special case, or whether Rawls' theory is strengthened by derivation from more powerful principles, is of little importance at this stage. In one respect, however, I think Rawls' account may be strengthened by what I have said here: the generality demanded by the requirement of rationality may be a justification, or perhaps a substitute, for the ignorance-clauses in Rawls' description of the "original position" from which the principles of justice are chosen. Indeed Kant can be seen as inserting another stage at the beginning of Rawls' four stage sequence of diminishing ignorance.(24) In the "original position" agents are imagined to know the general character of human capacities and interests, at the stage of the constitutional convention they know certain general facts about their society, and so on. Before the original position, I suggest, comes the noumenal stage, in which agents only know that they are rational agents, and what that entails. The various stages then form an ordered whole, bringing all of morality under the conception adumbrated by Rawls in his theory of justice: morality becomes rationality constrained by impartiality - morality as fairness.
1. This must have been written somewhere around 1972, with some notes added later. I
wouldn't write it the same way now.
2.H.J. Paton, The Moral Law, Hutchinson University Library, London, 1948, p. . (The pagination is that of the Prussian Academy edition of the Grundlegung).
3. Ibid, p. 
4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24.
4.Ibid, p. 
5.The Metaphysical Elements of Justice (Part I of The Metaphysics of Morals) tr. John Ladd, Bobbs-Merrill, p. 106. I once heard Keith Wrightson of Cambridge University explain why Kant could be expected to be sentimental about unwed mothers. It would be folly to try to reproduce the explanation now, in 1995, more than a decade afterward.
6.See D. Lyons. Forms and Limits of Utilitarianism, Oxford U.P., for references, and an account of the problem of relevant descriptions.
7.W.K. Frankena, Ethics, Prentice-Hall, p. 27.
8.The first approximation is derived almost wholly from lectures by David Lyons on the Grundlegung in 1966. I cannot be sure where I am following Lyons, and where not. But the basic outline is his. So I say - Lyons once denied it.
9.Lectures on Ethics, tr. L. Infield, Harper and Row, p. 44.
10.Paton, op. cit., p. .
11.C. D. MacNiven, "Kant on Perfect and Imperfect Duties", unpublished.
12.MacNiven does not, of course, hold that one ought to follow whatever conventions one encounters. He believes that Kant's theory is incomplete because it does not require moral assessment of conventions. In effect, what I see as a defect in MacNiven's interpretation of Kant, MacNiven sees as a defect in Kant's theory.
13.The similarity to Grice on meaning is not accidental. See H. P. Grice "Meaning", Philosophical Review, July 1957.
14.The analysis, if it is an analysis of anything, may be of something more general than promising.
15.A proper statement would be rather complex. See A. W. MacKenzie An Analysis of Purposive Behaviour, Ph.D. dissertation, Cornell, 1972; also P. M. Churchland "The Logical Character of Action-Explanation", Philosophical Review, April 1970.
16.The argument here is analogous - distantly - to that in Lyons, op. cit.. chapter three.
17.Chapter one of the Grundlegung suggests that Kant believes that universalizability follows from rationality. His "arguments" do not establish the connection, however. An interesting contemporary attempt to make the connection is found in T. Nagel, The Possibility of Altruism, Oxford, 1970.
18."Justice and Fairness", Philosophical Review, April 1958. Also A Theory of Justice, Harvard U. P. 1971.
19.Since a duty not to interfere arises without reference to possible conventional arrangements in a society, there is a fairly clear sense in which one can say that one has a natural right not to be interfered with.
20.This jibes nicely with Locke's belief in a natural right to property in "An Essay Concerning ... Civil Government", para. 27: "For this labour being the unquestionable property of the labourer, no man but he can have a right to what that is once jointed to, at least where there is enough, and as good left in common for others." It would be a mistake to erect this into a general theory of property, of course.
21.A possible source for a radical egalitarianism is the belief that the only reason for greater efficiency with unequal shares is a kind of trahison des doués. (This note appears to look forward to my "An Alternative Derivation of the Difference Principle", Dialogue, XIII, 4, December 1974.)
22.Frankena, op. cit., p. 27.
23.A practice of this kind would be just if it were modelled on a fairly designed program of insurance. One's readiness to give aid would be a premium paid against possible future needs.
24.A Theory of Justice, pp. 195-201.