The Unbreakable Law: Kant and the Categorical Imperative

Deontological ethics, from the Greek deon, meaning duty, holds that the morality of an action depends on whether it conforms to a rule or a duty, and not on its consequences. Its most influential exponent is Immanuel Kant (1724 to 1804), who argued that morality is grounded in pure practical reason: the moral law is not discovered through experience but is a requirement of rational agency itself.

Our research file names three core features of the position. First, agent-centered constraints: some actions are wrong regardless of how good their consequences might be, so murdering one person to save five is forbidden, torturing an innocent person to prevent a catastrophe is forbidden, and lying is forbidden even where lying would produce the better outcome. Second, moral rules bind categorically: not a conditional of the form if X is wanted then do Y, which is a hypothetical imperative, but a flat do Y, which is a categorical one, so that moral obligations are unconditional. Third, the right has priority over the good: what we are obligated to do constrains the pursuit of the best state of affairs rather than being derived from it.

Deontology is one of the three standing families of normative ethics, alongside consequentialism and virtue ethics, and it is not a historical curiosity: the Kantian line is actively worked today.

The Groundwork opens with the claim that nothing anywhere can be called good without limitation except a good will. The sentence our research file prints for it reads: "There is nothing it is possible to think of anywhere in the world, or indeed anything at all outside it, that can be held to be good without limitation, excepting only a good will." That English is Allen Wood's rendering of the passage at Ak 4:393, and not Mary Gregor's, although Gregor is the only translator our file's bibliography names.

Kant's own illustration is a shopkeeper. One shopkeeper gives correct change because he fears losing customers: he acts in conformity with duty, but not from it. Another gives correct change out of respect for honesty, even where cheating would be profitable: he acts from duty. On Kant's account only the second action has moral worth.

Kant argued that moral principles cannot be derived from experience, because empirical observation tells us what is and not what ought to be, and cannot be derived from human nature either, because the moral law must apply to all rational beings and not only to humans. Morality must therefore be grounded in pure practical reason: the same reason that enables us to think logically also legislates moral requirements.

For Kant the moral law is not imposed from outside, neither by divine command nor by social convention. It is self-legislated: rational agents impose it upon themselves through their own reason. This is autonomy, meaning self-legislation, and our research file names it the fundamental moral capacity and the ground of human dignity. Autonomy in turn requires freedom. Kant argued that morality presupposes free will, so that we must regard ourselves as free even if the phenomenal world is deterministic, which connects his ethics to the phenomena and noumena distinction of the Critique of Pure Reason.

A hypothetical imperative has a conditional form: if X is wanted, then do Y. It rests on a desire, our file's example being that a person who wants to pass the exam should study, and it binds only those who happen to have the relevant desire. A categorical imperative has the form: do Y. It is unconditional and binds all rational agents regardless of desires or circumstances. That, for Kant, is the form the moral law has.

The first and most famous statement of that imperative is the Formula of Universal Law. Mary Gregor's translation renders it, at Groundwork Ak 4:421, as: "act only in accordance with that maxim through which you can at the same time will that it become a universal law."

The test our file sets out runs in three steps. Formulate the maxim of the action: I will do X in circumstance Y to achieve Z. Imagine that maxim as a universal law, with everyone in that circumstance acting the same way. The action is morally impermissible if the universalized maxim involves either a contradiction in conception, where the universal law would make the action logically impossible, our file's example being that universal false promising would destroy the institution of promising, or a contradiction in will, where the universal law could not rationally be willed, our file's example being that nobody could rationally will that no one ever helps anyone, since anyone might need help.

The Formula of Humanity, in Gregor's translation, reads: "so act that you use humanity ... always at the same time as an end, never merely as a means." Our own research file prints it in a more modernised English, close to Thomas Kingsmill Abbott's rendering, and names no translator for it. The sense is not in dispute; the wording is not one single thing, and neither version should be left floating without a name attached.

Persons, in this framework, are rational agents with dignity, Würde, an unconditional and incomparable worth. Dignity is set against price, Preis: a thing with a price can be exchanged for an equivalent, and a person cannot. To use someone merely as a means is to treat them as a tool for one's own purposes, without respecting their rational agency.

The word merely is load-bearing, and our own file guards it explicitly. Kant does not prohibit all instrumental relationships; our file's example is using a taxi driver to get somewhere. What he prohibits is treating persons solely as instruments, without their rational consent or any regard for their own ends.

A third formulation, which our file gives as the Formula of the Kingdom of Ends, asks that one act only on maxims that a universally legislating member of a merely possible kingdom of ends could adopt. That sentence is given here as paraphrase, and marked as paraphrase, because no translator is named anywhere in our file for the wording it prints. A kingdom of ends is an ideal community of rational agents, each treating every other as an end in themselves and co-legislating moral principles, and our file notes that it anticipates the contractualist approaches of Rawls and Scanlon.

How many formulations there are is itself unsettled. Our research file names three, the Formula of Universal Law, the Formula of Humanity and the Formula of the Kingdom of Ends, under a heading that says three. The Stanford Encyclopedia of Philosophy entry on Kant's moral philosophy, in its substantive revision of 2 October 2025, names four, adding an Autonomy Formula that our file discusses at length under autonomy without ever calling it a formulation. Counts of three, four and five all appear in the literature depending on how the Formula of the Law of Nature and the Autonomy Formula are handled. This article therefore says several and states no number.

Kant divides duties into two kinds, and the division decides the hard cases. Perfect duties are strict and admit no exception. They are duties to refrain: no lying, no false promising, no suicide, no using others merely as means. Violating one is always wrong. Imperfect duties are wide and allow latitude in execution. They are duties to pursue: develop one's talents, help others in need. They must be adopted as ongoing maxims, but there is discretion in when and how, so nobody is required to help everyone at every moment, only to make beneficence part of a practical orientation.

Our research file states that Kantian ethics provides a powerful foundation for human rights, because the inherent dignity of rational agents is unconditional and cannot be overridden by any calculus of consequences, and that this underlies modern human rights discourse, the Geneva Conventions and medical ethics principles such as informed consent. The lineage is real and widely accepted. The causal version of it, that Kant's philosophy produced those instruments, is a much stronger historical claim than our file sources, so this article carries the point as intellectual descent and not as cause.

The same principle produces a conclusion harsher than modern readers expect. On Kant's account even punishment must respect dignity, and retributive justice, meaning proportional punishment for a crime, does so by treating the criminal as a rational agent who chose to act. Deterrence-only theories, which punish one person in order to influence the behaviour of others, treat criminals as mere means to social ends.

In On a Supposed Right to Lie from Philanthropy (1797), Kant argued that lying is always wrong, including lying to a murderer who asks whether a friend is hiding in the house. The duty of truthfulness, on his account in that essay, admits no exceptions. Our research file calls it his most controversial application.

Our file gives his justification in two parts. The first is about responsibility: a person who lies takes on responsibility for the consequences of the lie, whereas a person who tells the truth leaves the bad outcome as the murderer's responsibility and not their own. The second is the universalizability argument: universal lying would destroy communication itself, which our file describes as a performative contradiction.

Our research file records that the position has been criticized even by sympathetic Kantians, and that Korsgaard and others argue Kant's reasoning here is flawed and that his own principles could justify deception in extreme cases. It names no paper, no year and no argument for that, which is not enough to write from, so the two specific attempts below were found outside our file.

Christine M. Korsgaard, in The Right to Lie: Kant on Dealing with Evil (Philosophy & Public Affairs 15, no. 4, Autumn 1986, pages 325 to 349), argues that two of the formulations give different answers in this case. Under the Formula of Universal Law the lie to the murderer can be shown permissible, because the maxim of lying to a deceiver is universalizable, given that a murderer who conceals his purpose is himself relying on deception. Under the Formula of Humanity the same lie is impermissible, because on that formula coercion and deception are the most fundamental forms of wrongdoing. Her resolution is a distinction between ideal and non-ideal conditions: where strict compliance with the ideal embodied in the Formula of Humanity would make agents the tools of evil, agents should turn to the Formula of Universal Law for guidance.

Helga Varden, in Kant and Lying to the Murderer at the Door . . . One More Time: Kant's Legal Philosophy and Lies to Murderers and Nazis (Journal of Social Philosophy 41, no. 4, 2010, pages 403 to 421), argues from Kant's legal philosophy, as distinct from his ethics, that the standard reading of the essay is mistaken and that Kant's own position does not commit him to telling the murderer the truth.

The empty formalism objection, which our file attributes to Hegel and Mill, holds that the categorical imperative is merely formal: it says be consistent and supplies no content, almost any maxim can be universalized if it is described specifically enough, and it therefore generates no determinate obligations. The reply our file gives is that the universalizability test is not merely about logical consistency but about whether a maxim can rationally be willed as a universal law, which brings in human needs, vulnerabilities and the social conditions of action. Our file adds that O'Neill has shown the test has real moral content; that is her argued position rather than a settled result, and this article carries it as such.

Hegel's version of the objection appears in his Natural Law essay of 1802 and in the Philosophy of Right of 1821, where at paragraph 135 he objects that if the definition of duty is taken to be the absence of contradiction, then no transition is possible to the specification of particular duties. The paragraph attribution here follows the Hegel scholarship, Sally Sedgwick's chapter in A Companion to Hegel among it, and not a reading of the Philosophy of Right itself.

Mill's version is blunter. In chapter 1 of Utilitarianism, serialised in 1861 and published as a book in 1863, he states Kant's principle in his own paraphrase, "So act, that the rule on which thou actest would admit of being adopted as a law by all rational beings", which is Mill's rendering of Kant rather than a translation of Kant's German, and then charges that Kant cannot show any contradiction at all in the universal adoption of outrageously immoral rules.

The motivation objection, which our file attributes to Williams and Stocker, holds that acting from duty is cold and impersonal. Our file's illustration: a friend who visits a patient in hospital and explains that he came because it was his duty, rather than because he cared, has said something alienating, and Kantian morality seems to require suppressing natural affection. The reply our file gives is that Kant distinguishes acting from duty from acting merely from inclination and does not require the absence of feeling, only that duty and not feeling alone be the determining ground of the action, so that one can act from duty and with affection at the same time. Baron and Herman have developed sympathetic accounts of Kantian moral motivation.

The rigidity objection holds that Kantian absolutism produces absurd results: the truth must be told to the murderer, no lie is ever permitted, no promise may ever be broken, while real moral life requires sensitivity to circumstance. The reply our file gives is that many contemporary Kantians, Korsgaard, Herman and O'Neill among them, argue that Kant himself is misinterpreted, because the universalizability test requires careful attention to the maxim's description, so that a maxim of lying to protect an innocent life from a murderer may be universalizable; and that Ross's prima facie framework resolves conflicts without absolutism.

The problem of conflicting duties holds that when duties collide, keeping a promise against saving a life being the standard case, strict Kantianism supplies no resolution, because it insists that all duties are absolute. The replies our file gives are Ross's prima facie duties, which it calls the standard deontological resolution, and the Kantian constructivist claim that the categorical imperative itself can resolve conflicts through careful maxim formulation.

A fifth objection sits outside all of that. G. E. M. Anscombe's Modern Moral Philosophy (Philosophy 33, no. 124, 1958, pages 1 to 19), the founding document of the twentieth-century virtue-ethics revival, attacks the whole family of duty-based moral theories, Kant's included, on the ground that a notion of moral obligation and moral law is unintelligible outside the framework of a law-giver that modern moral philosophy has abandoned. Kant's own answer, that the agent legislates for himself, she treats as incoherent. That summary is the standard reading of the paper, assembled from secondary accounts rather than from the paper itself, and nothing here is quoted from it.

W. D. Ross, in The Right and the Good (1930), rejected utilitarianism on the ground that consequences are not all that matters, and rejected strict Kantianism on the ground that duties cannot be absolute because they sometimes conflict. In place of both he proposed prima facie duties: duties that are binding unless overridden by a stronger duty in the particular situation. He named seven. Fidelity, keeping promises. Reparation, making amends for past wrongs. Gratitude, returning kindness. Justice, distributing benefits and burdens fairly. Beneficence, improving the condition of others. Self-improvement, developing one's own virtue and knowledge. And non-maleficence, not harming others, which he took to be generally stronger than beneficence.

On Ross's account, when duties conflict the agent must exercise judgment rather than follow a formula, to determine which duty is stronger in the particular case. Our research file records that Ross acknowledged this introduces an element of moral intuition and does not provide an algorithm.

Charles Fried, in Contract as Promise (1981), gave the autonomy-based account: a promise is a moral act by which the promisor creates in the promisee a right to the promised performance, and its moral basis is respect for autonomy, the promisor exercising a capacity for self-binding commitment. The deliberate invocation of the promissory convention creates a genuine obligation, and Fried argued that the law of contract should be understood as the legal enforcement of voluntary promises. Our file names two critics, Atiyah (1981) and Craswell (1989), who argued that not all contracts are promise-like, since many are bargains, risk allocations or reliance-based arrangements that do not fit the promissory model.

T. M. Scanlon, in What We Owe to Each Other (1998, chapter 7), gave the expectations-based account: the wrongness of breaking a promise derives not from a mysterious power to create obligations through words but from the principle of fidelity to expectations. If I deliberately create an expectation that I will do something, another person relies on it, and I know this, then it would be wrong for me to disappoint that expectation without adequate reason. The account is contractualist: the moral force of promises derives from principles no one could reasonably reject. Our file records the criticism that this is too thin, making promise-breaking wrong only because of harm to expectations rather than any intrinsic wrong in failing to keep one's word, and Kimel's reply in From Promise to Contract (2003) that promises are also valuable for the intimate, trust-based relationships they express and sustain.

David Hume, in Book III of A Treatise of Human Nature (1740), gave the convention-based account: promises are not natural acts but conventional ones, dependent on the existence of a social practice of promising, so that without the convention the mere utterance of words would create no obligation. Their moral force derives from the social utility of the convention. That is the empiricist alternative Kant was writing against.

John Rawls, in Two Concepts of Rules (Philosophical Review, 1955), distinguished justifying a practice from justifying actions within a practice. The practice of promising can be justified on utilitarian grounds, because promising is useful; an individual promise must nevertheless be kept because the practice requires it. Once the practice exists, individual obligations are not reducible to case-by-case calculations of utility.

A fourth line describes the mechanism rather than the obligation. Austin, in How to Do Things with Words (1962), and Searle, in Speech Acts (1969), analysed promising as a performative speech act: the utterance I promise does not describe a pre-existing state of affairs but creates an obligation through the act of speaking. Our file gives Searle's conditions for a valid promise as the speaker intending to do what they promise, the hearer understanding the promise, the promise creating an obligation that would not otherwise exist, and the speaker being sincere, and adds that this linguistic analysis clarifies the mechanism of promising without by itself explaining its moral force.

Ross treated fidelity, meaning promise-keeping, as a prima facie duty: a real moral obligation that can nevertheless be overridden by stronger moral considerations such as preventing serious harm, justice or beneficence. Our file's worked example is a promise to meet for lunch, a serious car accident encountered on the way, and a duty of beneficence that overrides the promise. The promise is not cancelled by being overridden. It generates a residual obligation, so the right response is to apologise, explain and reschedule. Our file sets limits on this: trivial convenience does not override a promise, the competing obligation must be genuinely compelling, and the chronic promise-breaker who always has more important things to do undermines trust and social cooperation.

Christine Korsgaard's The Sources of Normativity (Cambridge University Press, 1996) locates moral obligation in the structure of practical reason and in what she calls practical identity: an agent needs to act on principles it can endorse on reflection, and when it reflects on its motivations it must be able to endorse them as a self-legislating agent, so that morality is the condition of rational agency itself. Her constructivism holds that moral truths are not discovered but constructed through the activity of practical reason, and that the categorical imperative is the procedural principle by which rational agents generate specific moral obligations.

Onora O'Neill applied the universalizability test to questions of justice, poverty and institutional design, argued that Kant's ethics is not merely an individual morality but supplies principles for just institutions, and pressed the practical applicability of Kantian reasoning against the charge of excessive abstraction. Her Constructions of Reason (Cambridge University Press, 1989) is the work our file names.

T. M. Scanlon's contractualism, in What We Owe to Each Other (Harvard University Press, 1998), holds that an action is wrong if it would be disallowed by any set of principles that no one could reasonably reject as a basis for general agreement. This is not about maximizing welfare but about finding principles each person can accept as a free and equal participant. It is Kantian in spirit, respecting persons as rational agents and seeking universalizable principles, and it differs from Kant in being explicitly interpersonal rather than derived from individual practical reason.

Derek Parfit, in On What Matters (Oxford University Press, 2011), argued for what he called the Triple Theory: that the best Kantian principle, the best version of rule consequentialism and the best version of Scanlonian contractualism all agree about which actions are wrong. His image, which gave the book its original working title, is of climbers ascending the same mountain by different routes and finding at the summit that they were heading for the same place. In outline, the Triple Theory holds that an act is wrong just when such acts are disallowed by principles that are optimific, uniquely universally willable, and not reasonably rejectable. That formulation is given here as paraphrase and not as a quotation from the book.

Our file prints the Formula of Universal Law as "Act only according to that maxim whereby you can at the same time will that it should become a universal law of nature." That string is a blend of two different formulations. Gregor's Formula of Universal Law at Ak 4:421 contains no words of nature; those words belong to the separate Formula of the Law of Nature, which Kant states immediately afterwards. Section 04 quotes the shorter form, which is also the form this wing's article The First Scales already quotes, so the two pages agree about the most famous sentence in the argument.

The good-will sentence our file quotes is Allen Wood's English rendering of Ak 4:393, while our file's bibliography credits the Groundwork to Mary Gregor's translation and names no other translator anywhere in the document. A reader would reasonably take the quotation for Gregor's. Section 02 attributes it to Wood.