Kantian Maxims and Lying
Alexander R. Pruss
November 17, 2000
Kant has claimed that lying is always wrong, even in response to a question from a murderer about the whereabouts of his intended victim. Christine Korsgaard has argued that although Kant’s second and third formulations in terms of respect for the humanity in persons and in terms of the Kingdom of Ends of the Categorical Imperative (CI) commit him to this claim, the first formulation in terms of universalizability does not in general. She has used this observation to argue first of all that the three versions of the CI are inequivalent and secondly that one could build a broadly Kantian two-level moral theory which has separate provisions for how to deal with the evil of others. In such a theory, the first formulation of the CI applies always, while the second and third formulations need not be applied in some cases of dealing with evil.
I shall argue that Kant’s first form of the CI does commit him to the claim that lying to murderers is wrong, for a very interesting reason having to do with the logical form of Kantian maxims. This observation of course removes the primary motivating example for Korsgaard’s two-level Kantianism. If one wishes to handle conversations with murderers looking for victims, one needs to look elsewhere, but I shall not discuss in this paper where. However, more interesting may be the point that I shall make about Kantian maxims and their objectivity. I shall argue that a paradigmatic Kantian maxim is the objective “When you are in circumstances C, do F in order to achieve E” rather than the subjectivized “When you believe you are in circumstances C, do F in order to achieve E.”
The first form of the CI (CI-1) states that one must not act on that maxim which one could not without contradiction simultaneously will to be a universal law. There are two ways a maxim can violate CI-1: it could be that one cannot conceive of a maxim as a universal law, in which case Kant says that acting on the maxim would be a violation of a “perfect” duty, or it could be just that one cannot conceive of willing the maxim to be a universal law, in which case it is a violation of an “imperfect” duty.
It is only the case of perfect duty that will be relevant to the question of lying. An issue much discussed in the literature has been what kind of a “contradiction” Kant is talking about in CI-1. I shall follow Korsgaard in assuming that Kant is talking of a “practical” contradiction: “the universalized maxim contradicts itself when the efficacy of the action as a method of achieving its purpose would be undermined by its universal practice.” The best example may be Kant’s: If universally practiced, the maxim to get money by falsely promising to pay it back whenever one needs money would undermine its own purpose, since no one would take such promises seriously.
One might think that for similar reasons lying is always wrong. But since most murderers do not go around saying they are murderers, Korsgaard assumes that the murderer believes that you, his interlocutor, do not know that he wants to murder the person for whose whereabouts he asks, but that in fact you somehow have found out about this. In other words, the murderer is trying to deceive you about his intentions. If saying a falsehood about the victim’s whereabouts in these circumstances is what we call “lying to a murderer”, then Korsgaard thinks that the maxim to lie to murderers is one that can without practical contradiction be conceived of as a universal law:
The lie will be efficacious when universally practiced. But the reason it will be efficacious is rather odd: it is because the murderer supposes you do not know what circumstances you are in—that is, that you do not know you are addressing a murderer—and so does not conclude from the fact that people in those circumstances always lie that you will lie.
This argument, however, is wrong. Consider the following categorical syllogism:
1. All people in those circumstances, i.e., those of addressing a murderer who wants to know the whereabouts of his victim, lie.
2. You are in those circumstances.
3. Therefore, you will lie.
If the maxim to lie to murderers were to become universal law, then almost all murderers would know (1), and in virtue of intending to murder they would know (2), and hence would conclude (3) and the maxim would undermine itself (except in a few very exceptional cases, namely those of murderers who do not observe what the universal law is or who are incapable of following the categorical syllogism). Therefore, the maxim to lie to murderers violates CI-1, pace Korsgaard.
A literal reading of Korsgaard would in fact have her saying that the murderer she describes is an exceptionally irrational one that knows (1) and (2) but fails to conclude (3). However, it is more charitable to take Korsgaard to be equivocating on the word “circumstances”. When she says that you do not know what “circumstances” you are in, “circumstances” means “that you are addressing a murderer”, while when she discusses the “fact that people in those circumstances always lie” by “circumstances” she means “when they believe they are addressing a murderer.” It is therefore crucial to Korsgaard’s argument that what should become a universal law is not the maxim “lie when you are addressing a murderer” but the subjectivized maxim “lie when you believe you are addressing a murderer.”
I shall argue, however, that in fact the maxim that the liar acts on is the non-subjectized “lie when you are addressing a murderer”, which maxim we have already seen is not universalizable. To do this, however, I first shall consider whether in fact we should not take all maxims to be subjectivized.
One might think that all maxims, when fully spelled out, will be in the subjectivized form “When you believe you are in circumstances C, do F in order to achieve E” rather than in the objective form “When you are in circumstances C, do F in order to achieve E.” After all, it is arguably the subjectivized forms that enter into explanations of actions. It seems that one explains Smith’s Fing not by saying that he was in circumstances C and acted on the objective form of the maxim, but by saying that he took himself to be in C and acted on the subjectivized form of the maxim.
However, this may be a misunderstanding of how reasons for actions work. Smith does not reason “I believe I am in C, and when I believe I am in C, I should F” but he reasons simply: “I am in C, and when I am in C, I should F.” The correct explanation of Smith’s Fing then is arguably that he believed he was in C and acted on the objective form of the maxim as best he knew. One might think the question on which explanation is correct is a merely verbal one, and the fact that one form of words would be preferred by Smith in his reasoning has no philosophical import. However, the considerations of universalization and lying to murderers show that the issue is not just one of language.
So, when Kant talks of maxims, is he talking of ones in objective form or of subjectivized ones? Which action-explanation is a better one from a Kantian point of view? One may have a presumption in favor of the objective form flowing from Kant’s commitments to the impersonality of reason, but that is not by itself a sufficient argument. I shall argue in four other ways that the objective form fits better with Kant’s project.
First and foremost, observe that for Kant, rationality is exhibited in self-legislation of ends. This self-legislation of ends is in practice effected by willing to act on a maxim. Maxims embody the values and ends that persons set for themselves. However, it is the objective form of the maxim that embodies a person’s values. Consider a person who values learning from the wise. This valuation is embodied in the objective form of the maxim: “Learn from those persons who are wise”, rather than in the subjectivized form: “Learn from those persons who you think are wise.” There is nothing particularly valuable in learning from those whom one thinks to be wise, unless it should happen that they are wise. It is the objective form that much more clearly reflects one’s values, and in light of the close tie between maxims and the value-setting role of Kantian reason, the objective form is more likely the right one.
Secondly, observe that in fact the objective form is more general than the subjective. One can, after all, consider the circumstances C1 of believing that one is in C, and then the objective form of the maxim involving circumstances C1 is equivalent to the subjectivized form involving circumstances C. This greater generality means that if we stick to the objective form of maxims we can make some distinctions that we could not otherwise make. Thus, a person could choose to legislate for herself the value of showing gratitude either more narrowly to those who did in fact benefit her, or more broadly to those who she believes benefited her. The first value might be motivated by a desire to repay those who have benefited her, while the second might be motivated by a more selfish desire to be seen by others and self as being the sort of person who shows gratitude to those who she thinks have benefited her. These are different values, and if we take the more general objective form of maxims, they lead to different maxims, namely “Show gratitude to those who benefited you” and “Show gratitude to those who you believe benefited you.” Of course, the two maxims collapse extensionally: any action that can flow from one maxim can also flow from the other. But because they embody different values and different intentions, they are intensionally different. The only way the proponent of subjectivized maxims could show the difference would by the maxims “Show gratitude to those who you believe benefited you” and “Show gratitude to those persons who have the property that you believe that you believe that they have benefited you.” However, these two subjectivized maxims, while perhaps intensionally different should one think that there is a difference between believing that p and believing that one believes that p, fail to embody a difference in intentions. Consequently, employing the objective forms lets one make distinctions between intentions that one could not with the subjectivized forms.
Another example where a distinction might be made would be the following two maxims: “If you believe p is true, then make all your beliefs consistent with the belief that p” and “If p is true, then make all your beliefs consistent with the belief that p.” The defender of the subjectivized form must claim that when someone says she was acting on the objective form, she in fact was acting on the subjectivized form. However, while the objective form of the maxim expresses the value of keeping all one’s beliefs consistent with truth, the subjectivized form only expresses the value of keeping all one’s beliefs consistent. Moreover, there is a clear difference between universalizing the subjectivized and the objective forms. Someone who has successfully followed a universalization of the subjectivized form would have tossed out both members of every inconsistent pair of beliefs she holds, while someone who has successfully followed a universalization of the objective form would only have tossed out the false member or members of such pairs.
Thirdly, Kant thinks that we have an imperfect duty to help other persons in their actions. Such help involves taking their ends to be ours. Since the ends are embodied in maxims, presumably this means that one should then help people to act out their maxims. But on Kantian grounds of autonomy, it is only the objective forms of the maxims that it is appropriate for other people to help one with. To see this, consider a rich philanthropist who values scientific research and wants to give financial aid to scientists who are doing worthwhile research, but who does not know much about science or about how to distinguish worthwhile research from quackery. On the subjectivized account of maxims, the philanthropist acts on the maxim: “If you believe someone is a scientist doing worthwhile research, then help her financially.” But what maxim would we adopt to help the philanthropist? There two options:
(A) Help bring it about that the philanthropist succeeds on acting on the maxim that if the philanthropist believes someone is a scientist doing worthwhile research, then the philanthropist helps her financially;
(B) Help bring it about that if you believe someone is a scientist doing worthwhile research, then the philanthropist helps her financially.
Option (A), though it is the most natural rendering of “helping the philanthropist act on her maxim”, is unacceptable. For while on Kantian grounds the fact that someone is following out a maxim gives us some prima facie reason for helping her, we certainly have no prima facie reason for adhering to (A), except insofar as we believe the philanthropist’s judgment to be accurate. Moreover, if (A) is the embodiment of the maxim to help the philanthropist, then helping the philanthropist will not involve correcting her errors as to who is really a scientist doing worthwhile research, since (A) does not call for such correcting. However, helping someone is supposed to be a sign of respect for the other’s rationality, and if the philanthropist wants to help people who really are scientists doing worthwhile research, then one has acted against her end insofar as one has helped her support quackery.
Nor is option (B) acceptable to a Kantian, as it involves paternalism, which is certainly not what helping another person in her activities is supposed involve. On option (B), it is sufficient for you to choose who is a scientist doing worthwhile research and then to bring it about that the philanthropist helps this person financially. Doing this does not show Kantian respect for the philanthropist’s humanity, since (B) as it stands does not involve even convincing the philanthropist that the person you think is worthy of help is worthy of help, but simply trying to get the philanthropist to help these people, conceivably by convincing the philanthropist that they are for other reasons worthy of help, which again is not what helping the philanthropist in her self-appointed maxim involves.
On the other hand, the philanthropist’s maxim is in the objective form “if someone is a scientist doing worthwhile research, then help her financially”, the maxim to help the philanthropist can be:
(C) Help bring it about that the philanthropist fully succeeds in acting on the maxim that if someone is a scientist doing worthwhile research, then the philanthropist helps her financially.
This maxim escapes from the objections to either (A) or (B). For if the philanthropist helps a quack, she does not succeed in acting on her maxim, though she may have sincerely tried to do so. Hence, adopting (C) as one’s maxim gives one prima facie reason to prevent the philanthropist from helping quacks financially. However, adopting (C) avoids the paternalism in (B). For if we simply find someone we think is a scientist doing worthwhile research and convince the philanthropist to help this person, we have not helped the philanthropist to act on her maxim, since in helping this person the philanthropist is not acting on her maxim. Rather, we need to convince (and for other Kantian reasons, in a way that respects the philanthropist’s rationality) the philanthropist that the prospect for her help is a scientist doing worthwhile research. Only then will we have helped her act on her maxim.
Finally, let us imagine that drug A cures an unpleasant but non-deadly disease D while killing people who have a more rare, though not rare to the point of negligibility, disease E, which is also unpleasant but non-fatal. Suppose further that cases of disease E are very easy to mistake for cases of D, even by well-trained physicians. Now, consider the following maxim:
The action contained in the maxim is morally acceptable. If a patient is known to have D, then prescribing A is the right thing to do. However, the maxim in question is one that cannot be universalized when subjectivized. For the subjectivized form of the maxim is:
But were everyone to act on this maxim, the lot humankind would grow less happy, since many doctors will think they know that the patient has D whereas in fact the patient has E which makes the medication fatal, and the alleviation of the unpleasant symptoms of D produces less happiness overall than the mourning caused by the deaths. Thus, were we to generalize maxim (E), we would defeat the purpose behind (E), thereby involving ourselves in a contradiction in conception.
Hence, (E) fails the universalization test. But giving A to a patient one knows to have D is morally acceptable. Thus, to check the acceptability of (D), we should not subjectivize it to (E).
But perhaps the action described in (D) is immoral. For, if the doctor knows that E is often misdiagnosed as D, then perhaps she should not prescribe A even when in fact she knows that the disease is D, since for aught that she knows, she may not be in a position of knowing that the disease is D. This objection presupposes that one may know that p without knowing that one knows that p. Let us assume this principle for the purposes of discussion. Then, I can redo the argument by considering the following two maxims:
(F) If you are a doctor, and x is a patient such that you know that you know that x has D, then prescribe A in order to make the lot of humankind happier.
Surely, if someone not only knows that x has D but also knows that she knows it, then prescribing A is acceptable. But it might still be the case that were every doctor who merely thinks that she knows that she knows that a patient has D to prescribe A, on the whole people would be less happy, and so maxim (G) would be self-defeating if universalized as its purpose would not be achieved. Hence, once again, subjectivizing a maxim prior to universalizing gives the wrong answer about the rightness of the act.
Therefore considerations on what is involved in helping people successfully act on their maxims are yet another way of seeing that Kantian maxims are primarily in objective form, though in special circumstances of course the objective description of circumstances may involve propositions about the agent’s beliefs.
I have argued then that the primary form of a Kantian maxim is the objective. Thus the person lying to the murderer should be taken to be acting on the objective form maxim “Lie to murderers” rather than on the maxim “Lie to people you think are murderers.” But now there is a difficulty, in that it is unclear how the universalization is supposed to work. After all, in the imaginary world where everyone successfully acts on the objective form maxim, people will end up lying to all murderers even if they do not know them to be murderers. But that is absurd, surely, because in such cases they would not be acting on the maxim, since acting on the maxim requires that one believe one’s interlocutor is a murderer. So why would these people be lying to the murderers if they did not know them to be such? (Of course we must be careful here, since we are talking of a world which is ultimately impossible, at least “practically”, since the maxim in question fails the CI-1 as has been shown.)
Kant, however, has an answer here. The CI in the first form does not say that a maxim is unacceptable when it cannot be willed that everyone acts on the maxim, but rather when it cannot be willed that the maxim is a universal law. By “universal law” Kant here means a law of nature, and so the explanation of why the people in the imaginary world lie to murderers is not that they are acting on a maxim but the explanation is simply in terms of natural causality. Just as the good will is not to be conceived of as acting from duty, so too these people whom one has (imaginarily) willed to always infallibly lie to murderers do not act on a maxim.
Moreover, it is arguably in general impossible directly to see to it that another person should act on such-and-such a maxim (compare how in Belnap’s formalization it is impossible to see to it that someone else sees to it that p), and if one knows this is impossible, then one cannot even will it. If this is right, then if the imaginary world where everyone does what is described in the maxim were one where everyone acts on the maxim, then it would always be contradictory to will that things be so, and so every maxim would automatically fail the CI-1. So, to save the CI-1 from absurdity, one need not imagine that in the imaginary world people do not act on the maxim. It is enough to imagine that they do what the maxim says, for whatever reasons. And in the case at hand, we see that then we have a practical contradiction.
But, perhaps, someone could still choose to act not on the maxim lie to murderers which we have seen fails CI-1 but on Korsgaard’s lie to people you think are murderers. However it is not clear what it would be to act on that maxim. We cannot simply arbitrarily decide what maxim we are to act on. Otherwise CI-1 would be practically trivial, since any action becomes universalizable once its description becomes sufficiently specific. E.g., consider killing the noisy neighbor under the alleged maxim “Murder all men who play the piano loudly at night in third-story south-facing apartments, who are 181.223904 cm tall, who weight 77.5322 kg, and who were educated as accountants” which might well pick out a single individual. However, while this might have the grammatical form of a maxim, it would be very difficult to imagine how it could actually be a maxim of a sane human being, and hence of any human being since only the sane have maxims, since the height and weight of the noisy piano player is rationally irrelevant to the reasons one has for killing the noisy neighbor.
One cannot choose one’s maxim at will, and as we have seen in the previous section, the basic form of a maxim is the objective form. We can also see this directly in the case of lying to the murderer by applying the test of what it would take for me to help you successfully follow out your maxim. In doing so, I would not be just adopting the maxim to help bring it about that you lie to those that you think are murderers, since surely respect for the values embodied in this maxim would require that I help you figure out who is a murderer and who is not. Nor would I be adopting the paternalistic maxim to help bring it about that you lie to those that I think are murderers, since that might involve you in lying to people who you think are decent folk, and it would be a violation of your autonomy, a use of you as a mere tool, if I did that. Rather, to help you follow out your maxim, I would have to help to bring it about that you lie to those who are murderers, which would involve me helping you to decide who the murderers are and how most effectively to lie to them. Since this is what it takes to help you follow out your maxim, it must be that your maxim is in fact the maxim to lie to murderers simpliciter. That you think someone is a murderer will of course help explain why you think lying to him is in accord with the maxim. But the condition that you think he is a murderer is not built into the maxim’s conditions.
Maxims set by rationality are in general not subjectivized, and so lying does violate the first form of the Categorical Imperative. What implications one will take this to have for Kant’s project will depend on whether one thinks the absolute prohibition on lying is morally tenable. One might of course agree with Korsgaard that Kant’s second and third forms of the Categorical Imperative do give one an attractive reason for why lying wrong and, now that Korsgaard’s two-level approach has to be rejected, one might end up agreeing with Kant that lying is absolutely wrong. One might additionally try to use other methods to soften the prohibition, arguing that some things commonly taken to be lies are not. Or one might take the conclusion of this paper to be reason to reject Kantian ethics in general.
 For one possibility, see my “Lying and Speaking Your Interlocutor’s Language”, The Thomist ??ref.
 Korsgaard, p. 135.
 Korsgaard, p. 136.
 If one thinks that this judgment relies too much on contingent matters of fact about how often E is mistaken for D and the dangerousness of A, then one can modify the two maxims by adding the condition “and if E is a disease for sufferers from which A is fatal but which is easily mistaken for D and if D and E are unpleasant but non-fatal”.
 Of course it could now be objected that it is not enough to know that you know that x has D but you must know that you know that you know this. To cut short such objections, say that you “strongly know” that p if Knp for every n, where Kn is short for n iterations of the phrase “you know that”, and modify the two maxims by replacing “you know that you know that” by “you strongly know that”. It might still be the case that the happiness of humankind would be harmed if every doctor who thought she strongly knew that a patient had D were to prescribe A, but surely there would be nothing morally wrong with a doctor’s prescribing A when in fact she did strongly that the patient had D.
 More precisely, insofar as one is acting on a maxim, one is sane.
 I take that approach in Pruss (op. cit.)