Suppose a German SS officer knocked on my door, asking me whether I had any Jews. And suppose further that I had two Jews in a secret compartment in the attic that he'd never be able to find. Everybody will agree that I must lie and say I haven't any Jews in my house.

But I'd have to disobey Kant's categorical imperative "do not lie", because I felt obliged to "not betray innocent people leading to their death".

Both imperatives can be supported if I ask myself the question "what if everyone did it?". If everyone lied whenever they felt like it, society and civilization would collapse. Therefore I must never lie. But society would be a grim, inhumane place if everyone cooperated in killing the innocent. Therefore I must never betray those under my protection.

How can this be solved? I remember hearing two "solutions":

  1. I must heed both imperatives and choose a third option, thereby violating neither: "I refuse to answer you". This would result in my death, but I would have upheld both imperatives.

    I find this absurd; I do not think Kant would have wanted this. And yet, when challenged by Constant with a similar situation, he said that one should challenge the murderer or refuse to answer. In this case, that would lead to my house being burned down, the Jews dying an even more horrible death.

  2. The "do not lie" imperative is too broadly formulated: it should be "do not lie, except when your lie would save another's life". I find this very weak: the power of a categorical imperative would be that it is a broad statement. It should be a universal law.

    The fundamental problem would seem to be that each rule saying "I wouldn't want everyone to do this" has a reasonable exception, and that it should not be left to each single person to determine what the exceptions are depending on a particular situation. I believe Schopenhauer offered similar criticism.

How did Kant solve this problem while keeping his categorical imperatives intact? Or, if he didn't, how could it be solved hypothetically?

  • 1
    @Joe: I'd want some kind of prioritizing as well; but then it seems we have to dump the universality of the imperatives...
    – Cerberus
    Commented Jun 10, 2011 at 2:17
  • 2
    Good question. The categorical imperative has always bothered me for this reason, it seems to point to solution 1, which seems to sacrifices the self (and others) for no greater reason than because if everyone did it all the time it would be bad. I'd love to hear what a Kant scholar has to say on the topic.
    – dimo414
    Commented Jun 10, 2011 at 7:17
  • 2
    Isn't this just another case of no finite number of rules can cover all the possibilities? If Kant was logical he would either Lie or not Lie, in either case he would end up breaking his own rules. What if the you lie to save the life of somebody going to take more lifes? that exception would need another excpetion.
    – jimjim
    Commented Jun 10, 2011 at 11:48
  • 2
    You have effectively broken do not lie when you choose to provide shelter and hide them where they could not be found.
    – Chad
    Commented Jun 10, 2011 at 13:54
  • 3
    An interesting point @Chad, so to a Kantian, even putting yourself in a situation where your actions are not entirely upfront and clear is a violation of the CI? It makes sense in a way, but seems to me makes the CI all that more of an utterly impossible ideal.
    – dimo414
    Commented Jun 10, 2011 at 22:47

8 Answers 8


Kant himself (and many other deontologists that follow in his footsteps) would never ask the question: "what if everyone did it?" In fact, it's quite irrelevant to their moral calculi. He famously decrees that:

The greatest violation of a human being’s duty to himself regarded merely as a moral being (the humanity in his own person) is the contrary of truthfulness, lying. In the doctrine of right an intentional untruth is called a lie only if it violates another's right; but in ethics, where no authorization is derived from harmlessness, it is clear of itself that no intentional untruth in the expression of one's thoughts can refuse this harsh name. [ ... ] By an external lie a human being makes himself an object of contempt in the eyes of others; by an internal lie he does what is still worse: he makes himself contemptible in his own eyes and violates the dignity of humanity in his own person. [...] By a lie a human being throws away and, as it were, annihilates his dignity as a human being.
(Kant, The Metaphysics of Morals, 1996, p 182)

As you observe, from a strictly Kantian point of view, there is absolutely no way that a lie can ever be justified. It is a complete violation of human agency, insofar as it manipulates someone (effectively treating them as a means rather than an end in and of themselves) into believing something for whatever reason. While it could be argued that the overall goal of the lie is a moral good, Kant would reject it a priori for its means, which he would consider sharply and incontestably immoral. That immorality outweighs and precludes any morality that might come later from the results of the action.

Thus, it makes absolutely no difference with respect to the morality of the original decision (i.e., whether or not to lie to the soldier) whether the eventual outcome of your decision is utter social chaos or collapse. That's an ends-based calculation, one that Kant says you can't possibly know at the time and shouldn't possibly matter when assessing the morality of one individual action.

At least one of the justifications that Kant would (and did) provide for this particular moral choice is that you are not responsible for the death of the two Jews, even if your telling the truth leads to their brutal murder by the Nazi soldiers. The consequence may be undeniably horrible, but it wasn't your crime and you weren't directly responsible for its execution. Essentially, he's saying, "that's on them, not you."

However, if you lie, you now take ownership and responsibility in the atrocity. Or, in other words, you become morally culpable for the immoral actions of the Nazi soldiers because you are essentially manipulating the soldiers yourself through your decision to lie and distort the truth. And once you've made the choice to distort the truth, you've established control over the situation and you should be held responsible for whatever immoral outcome may result. This is the infamous causing versus allowing distinction.

All of that to say that this is really not a contradiction or dilemma for a strict Kantian. Of course, that hardly means that it fails to be one for most of us, who take a far less extremist view of morality.

It's also worth considering that a neo-Kantian would take issue with your example on technical level, arguing that it's not a good example because the Holocaust is an exceptional situation, a special case, if you will.

More specifically, the German state at the time formed an illegitimate government, and the SS officers were acting as agents of this illegitimate government, this state that is on face unjust. Given such case, one could argue, his entire corpus of ethics does not apply, as it was only intended to apply to situations involving a just government. (For more on this, see Kant's "Doctrine of Right" and Metaphysics of Morals 6:264, etc.)

Finally, if a Kantian discussion of this exact question interests you, you might find this recently-published article to be a worthwhile read.

  • 6
    Excellently put, and very informative. Thank you.
    – dimo414
    Commented Jun 10, 2011 at 20:37
  • 2
    Thanks! It seems you and Mfg agree with Chad. You get the tick because you explained the problems and possible Kantian replies most completely. (My personal conclusion must be that Kant's potential answer would be poor. That consequences should not matter at all is simplistic; for how can it be determined whether an intention is evil if we disregard what results the agent expects? That categorical imperatives are suspended in certain societies would seem too complicating and arbitrary for a "universal" rule. But that must be why Constant and Schopenhauer found his moral philosophy lacking.)
    – Cerberus
    Commented Jun 11, 2011 at 2:30
  • @Cerberus: Those are far from the only two who found his moral philosophy to be quite lacking. As I observed, though the scenario you describe did not appear to present a strong dilemma to Kant himself, it is certainly one to the vast majority of us who take a less hard-line approach. Under a strictly deontological framework, there is indeed no room for a consideration of expected consequences. But schools of thought like virtue ethics and even David Cummiskey's neo-Kantianism, which attempts to combine elements of consequentialism with Kant's thought in his book Kantian Consequentialism. Commented Jun 11, 2011 at 5:35
  • I think no one can really solve the dilemma - because in reality you are never 100% what will be the outcome of your actions... If killing someone has a 51% chance of preventing the dead of two others, should you do it? In the end you will have no moral guideline and everything is just guesswork and possibly lying to yourself. - So if you say everything regarding the outcome is guesswork the only absolute thing left is our untainted direct action, this is the only thing we can control and should be perfect ;-)
    – Falco
    Commented Jul 21, 2014 at 10:45

While the question is an interesting one, there are several issues at hand to consider. First, the question seems premised on the belief that there can be multiple categorical imperatives. This is not Kant's view of the matter. For Kant, there can be multiple hypothetical imperatives but only one categorical imperative (not all contemporary Kantians agree). Hypothetical imperatives derive from either the goal of happiness (to be happy you need to eat) or the desire to accomplish some task (To make bread, you need yeast). But the Categorical Imperative follows from the dictates of reason.

I next want to look at this paragraph in your question:

Both imperatives can be supported if I ask myself the question "what if everyone did it?". If everyone lied whenever they felt like it, society and civilization would collapse. Therefore I must never lie. But society would be a grim, inhumane place if everyone cooperated in killing the innocent. Therefore I must never betray those under my protection.

This paragraph has several interesting features. First, there's a misunderstanding of how the categorical imperative works. Second, there is an important slippage in the language between the two universalizations.

On the first point, I will limit my remarks to the universalization model of the Categorical Imperative. The question is not "what if everyone did it?" but rather what if it was always to happen or in a possibly more loose way "what if it became a law of nature?" Moreover, the problem that is supposed to raise is not necessarily about societal collapse but about contradictions in implementation (I will always tell lies is a system that no longer works when every rational creature is always telling lies -- regardless of social consequences it eliminates the meaning of its own criterion.)

Second, the characterization of telling the man at the door about the people you are hiding is problematic. Specifically, when it is described as "cooperat[ing] in killing the innocent." This cannot possibly be an expression of the categorical imperative. But this is also not what Kant expects of the agent. Instead, Kant expects the agent to cooperate with the request of a seemingly rational agent (this distinction is crucial to Korsgaard's modification of Kant's answer). On Kant's account, we must assume the requests we receive from other humans no matter how unreasonable are founded on reason. In other words, despite the blood on the axe, we have to assume they are asking -- not to engage in the immoral act of killing the innocent but for some other reason.

Thus, for Kant, there's no conflict between these two ideas. In some ways, this makes things worse -- because it undermines one of the better outs we have on the picture where we cannot lie. To some extent, Kant gives us a positive duty to tell the truth to other rational beings -- not merely a negative duty not to lie. If we look in the Metaphysics of Morals, Kant does seem to have little leeway in suggesting we needn't purposelessly express the truth.

  • 1
    +1. This is clearly the best answer.
    – Geoffrey Thomas
    Commented Oct 28, 2018 at 11:24

I agree with what Chad says in that there are other options and that, to answer your question, civil disobedience (refusing to answer) suffices a solution to this particular problem. Consequentialism (your house getting burned down) is not a sustainable justification for disavowing a categorical imperative. This is because categorical imperatives deal in absolutes; as a result, an consequential analysis must have an absolute result, otherwise it fails.

At base however is whether any two categorical imperatives can conflict. (Assuming categorical imperatives can exist,) if two categorical imperatives (assuming that even one could exist) were to "conflict" would that mean that one was wrong, or perhaps that there was some temporal incompatibility that has yet to be resolved which upon resolution will show the two to be compatible?

The point being, assuming that two imperatives conflict, and we are dealing in some absolutist terms, and any solutions are mutually exclusive in meeting the needs of both; it is likelier that one categorical imperative is wrong or misunderstood than that no solution is useful.

  • Yeah, civil disobedience is an interesting point here. I considered addressing it, but then grew worried that my answer was far too long already. In particular, Kant allows for passive civil disobedience, although he specifically forbids any type of rebellion against the state. A "negative resistance," such as the refusal of the people to obey all of the laws is okay, but a full-out revolt is not. Citizens are only obligated to obey the sovereign "in whatever does not conflict with [their] inner morality". This is what I hinted at in the latter part of my answer, the "Doctrine of Right". Commented Jun 10, 2011 at 13:31
  • @Cody true. In the instance of this case, I was trying to remain abstract and got a bit further off; good job including both the less particular concerns of this example with the Kantian guidelines while keeping them pertinent
    – mfg
    Commented Jun 10, 2011 at 13:48
  • +1 Thanks! The argument that we must be "misinterpreting" one imperative or the other in case of an apparent conflict sounds authentic, from what I remember reading Kant, unsatisfactory though it may be
    – Cerberus
    Commented Jun 11, 2011 at 2:35
  • @Cerberus Yeah, plus it is shaded by my personal reading of Wittgenstein and the idea (i forget whose) that there are no disagreements, just misunderstandings (deliberate or otherwise)
    – mfg
    Commented Jun 11, 2011 at 16:32

Kant already responded to this same thought experiment.


ON A SUPPOSED RIGHT TO TELL LIES FROM BENEVOLENT MOTIVES. 1 - Immanuel Kant, Kant’s Critique of Practical Reason and Other Works on the Theory of Ethics [1785]

  • 2
    Thanks for that link! So did Kant really write that? Although that text propounds Kant's views, I read references like "the German philosopher", which was clearly Kant: would he refer to himself so?
    – Cerberus
    Commented Sep 25, 2012 at 0:16

You can choose not to answer. If he will never find them then by not answering then you are neither lying nor not betraying innocent people leading to their death.

Taking no action would be preferable to taking an action against the moral imperative.

Page 112 of this document shows that Kant clearly believed that lying to a murderer that someone is not there is a lie. Even if the person was not actually there but the actor believed he was.

However, a lie of omission is permissible. A declaration that is either known or believed to be false is not. However, if the Jews had been given refuge by the actor's wife but the actor did not see them it would be permissible to say "I have not seen any Jews this evening," inviting a conclusion that there are no Jews present, while not actually answering the question

  • You're probably saying what Kant said... but, in this case, disobedience would lead to my house being burned down, the Jews dying an even more horrible death. In general, saying "I won't say" usually makes it clear what your answer would be: "are you gay?" — "I'd rather not say" (you're gay); "do you like my dress?" — "I'd rather not say"/pause (you hate it), etc.
    – Cerberus
    Commented Jun 10, 2011 at 13:11
  • You do not know that. The SS Officer could choose to come in and do a search. Finding nothing leave with out causing any damage. Non-Compliance is not the same as disobedience. I did not say to answer that "I wont say" i said not to answer. Remain silent.
    – Chad
    Commented Jun 10, 2011 at 13:16
  • @Chad: Well, in the situation I have provided, he would have my house burned down if I remained silent. Let's not get into an historical discussion: my question is about what if... if you do not like this particular example, you can take another situation in which anything but a lie would mean disaster.
    – Cerberus
    Commented Jun 10, 2011 at 13:20
  • 2
    With regards to your clarification, it's worth considering that choosing not to act is itself a choice. That is to say, you can't avoid a moral dilemma merely by refusing to do anything at all. Refraining from taking an action is still taking an action, it's simply a different action altogether. But there are still all sorts of potential moral dilemmas attached, like tacit consent, complicity, civil obedience, etc. If only moral dilemmas could be as simple as "neither one!" Commented Jun 10, 2011 at 13:49
  • 1
    I agree and IF as in this example the Officer were to burn the house the moral imparitive would be to attempt to save those you chose to provide protection.
    – Chad
    Commented Jun 10, 2011 at 14:02

Actually, Corrie ten Boom recounts a story very similar to this in her book, The Hiding Place.

The following is not from the book itself, but from a website containing book notes for the text. It describes an incident just like this, where Corrie's younger sister, nicknamed "Cocky," refused to lie to Nazi soldiers:

One night, it happened in their neighborhood, and Peter and his older brother, Bob, rushed into Nollie’s house, looking for a place to hide. Nollie put them in her secret spot in the potato cellar under the kitchen table. When the soldiers burst down the door, they demanded the boys’ younger sister, Cocky, tell them where her brothers were. Without missing beat, she said they were under the table. When the soldiers lifted up the tablecloth, Cocky began to laugh, and so did everyone else. The soldiers, feeling humiliated, left, and the ten Booms spent the rest of the evening feeling both grateful for their safety and arguing over Cocky’s insistence on telling the truth. Nollie stood by her daughter and said, “God honors truth-telling with perfect protection!”

  • 2
    Does that anecdote have some bearing on Kantian ethics? I seem to be missing the connection. Commented Oct 26, 2011 at 18:03
  • @Mich: I think it means that God will reward honesty in some mysterious way even if it seems to betray the innocent at first. I highly doubt Kant's religion was so focused on providence.
    – Cerberus
    Commented Oct 26, 2011 at 18:28
  • @MichaelDorfman, I was addressing the second part of the question: "Or, if he didn't, how could it be solved hypothetically?" The incident quotes is an illustration of a possible solution of the exact scenario the questioner asked about.
    – JKubecki
    Commented Nov 2, 2011 at 17:15

Acording to the Catagorical imperative, the one wanting to kill the jews is breaking the rules, and your intent is for him to use his reason, to make an autonomus desicion. If all were to follow it, it would work, therfore you should. It only causes conflict when an autonomus agent dosent follow the catagorical imperative.

  • Welcome to the site! Can you explain what you mean by "it", exactly?
    – Cerberus
    Commented Jul 31, 2017 at 16:17
  • The question whether we should encourage moral behaviour is different from the question whether it can be allowed to lie in that situation with the Kantian CI as the criterion.
    – Philip Klöcking
    Commented Jul 31, 2017 at 20:15

"...it would be permissible to say " I have not seen any jews this evening." Inviting a conclusion that there are no jews present."

If it were known the Jewish people were in the home but not seen by the actor, to respond to the question in the way you've suggested would be acceptable would serve a dual purpose. 1 - It follows the duty to tell no lies, and 2 - It serves the purpose of treating the asylum seekers as an ends. Unfortunately by doing something that one hopes will deceive another the act is such that it treats those who are hopefully deceived as a mere means to ends, and I doubt Kant would accept attempts to deceive as a "good" universal law.

  • Unfortunately it is not clear that Kant thinks this option is plausible. The main issue is that he understands lies in an expansive way that would include dissembling.
    – virmaior
    Commented Apr 23, 2014 at 1:57

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .