Deontology is a family of theories in ethics according to which the rightness of an action does not depend solely on how good its consequences are. Some actions may be required, forbidden, or permitted because they fulfill duties, respect persons, or avoid violating rights, even when another action would produce a better overall outcome. A promise, for example, may give its maker a reason to perform a particular action that cannot be reduced to the benefits of promise-keeping in general.
Deontological theories differ over the source and strength of obligations. Some ground them in rational agency, others in rights, special relationships, or principles that people could reasonably accept. They also disagree about whether prohibitions admit exceptions, how conflicting duties should be resolved, and what moral significance attaches to intention, consent, and the distinction between doing harm and allowing it. Immanuel Kant provides the most influential systematic example, but deontology includes important non-Kantian approaches.
Deontology is usually contrasted with consequentialism, which makes consequences fundamental to determining right action, and with virtue ethics, which gives explanatory priority to character and human flourishing. These divisions concern the foundations of moral judgment rather than mutually exclusive lists of practical rules: all three approaches can condemn cruelty, support honesty, and recognize duties of assistance while explaining those judgments differently.
Scope and conceptual structure
The defining feature of deontology is not simply an interest in duties. Consequentialists also recognize duties, including duties to prevent suffering or promote welfare. The distinguishing issue is whether all obligations ultimately derive from an instruction to produce the best available outcome. Deontological theories deny that this maximizing requirement exhausts morality. An action may be wrong because it constitutes an assault, betrayal, or deception, rather than merely because it leaves the world worse overall.
Three elements often appear in deontological accounts: obligations, constraints, and permissions. An obligation requires an action, such as repaying a debt. A constraint limits what may be done in pursuit of an otherwise worthwhile end, such as prohibiting the deliberate killing of an innocent person to secure benefits for others. A permission leaves room for choice, including choices that do not maximize impersonal value. Some theories recognize a prerogative to devote substantial resources to personal projects rather than continually directing them toward the greatest aggregate benefit. These elements need not all have the same philosophical foundation.
A further distinction concerns the relationship between the right and the good. The claim that the right has priority over the good generally means that an independently specified account of valuable outcomes does not by itself determine permissible conduct. It does not mean that happiness, welfare, or other goods are unimportant. A theory can require substantial assistance to others while insisting that assistance must not be financed by theft from an uninvolved person. Nor does deontology necessarily deny that outcomes help determine which duty is most urgent.
Deontology is also distinct from moral absolutism. An absolutist prohibition admits no exceptions within its scope; a deontological prohibition need not be absolute. Some theories recognize defeasible duties, while others allow constraints to be overridden when sufficiently grave harm is at stake. Likewise, deontology is not identical with obedience to law, custom, or authority. A legal command may conflict with a moral duty, and a deontological theory may require resistance rather than compliance.
Finally, deontology is a position in normative ethics, not a complete theory of moral truth or knowledge. It does not by itself settle questions in metaethics, such as whether moral judgments describe objective facts or express practical commitments. Nor does it automatically determine how obligations become known. A rationalist account, an intuitionist account, and a contractualist account may all support deontological conclusions while disagreeing about their justification.
Historical formation and terminology
The word derives from the Greek deon, meaning what is binding or required, and logos, meaning an account or study. Jeremy Bentham’s posthumously published Deontology; or, The Science of Morality appeared in 1834 under the editorship of John Bowring. Its title did not designate the modern contrast between duty-based ethics and consequentialism: Bentham was a major exponent of utilitarianism. The history of the word must therefore be distinguished from the history of the theories now classified under it.
Accounts of duties long predate the modern terminology. Stoicism developed an influential vocabulary of appropriate conduct, rational nature, and obligations associated with human roles. Cicero’s De officiis, written in 44 BCE, examined duties connected with justice, social relationships, and honorable action. Such works supplied later European moral philosophy with important concepts, but describing them simply as modern deontology would conceal differences in their accounts of virtue, nature, and the human good.
Religious and natural-law traditions likewise developed prohibitions and obligations that were not treated as straightforward calculations of aggregate welfare. Thomas Aquinas connected moral precepts with practical reason and a teleological understanding of human nature. Divine-command theories instead locate obligation in divine commands, although their versions differ over the relationship between divine authority and goodness. Neither natural-law ethics nor religious ethics is automatically deontological: each contains diverse accounts of ends, virtues, rules, and moral authority.
Kant’s Groundwork of the Metaphysics of Morals (1785), Critique of Practical Reason (1788), and The Metaphysics of Morals (1797) established the most influential modern framework centered on rational moral law and autonomy. His account was not merely a catalogue of duties. It attempted to explain why moral requirements bind rational agents categorically, independently of contingent desires. Nineteenth-century debates subsequently placed Kantian ethics alongside utilitarian and other approaches, although the familiar textbook classification crystallized later.
Twentieth-century philosophy broadened the category substantially. W. D. Ross’s The Right and the Good (1930) defended a plurality of basic duties rather than derivation from a single Kantian principle. Later writers examined rights, moral constraints, responsibility, and the permissibility of harming one person to benefit others. Work by Elizabeth Anscombe, Philippa Foot, Judith Jarvis Thomson, Thomas Nagel, and others helped make these questions central to analytic ethics. Contemporary deontology consequently names a diverse research tradition rather than a single doctrine inherited unchanged from Kant.
Kantian duty, autonomy, and respect for persons
Kant distinguishes acting in accordance with duty from acting from duty. A merchant who charges fair prices solely to preserve a profitable reputation behaves in a way that may conform to duty, but that motive does not give the action the distinctive moral worth Kant associates with respect for moral law. The distinction concerns the basis of the agent’s choice. It should not be read as a simple claim that an action loses moral worth whenever affection, sympathy, or satisfaction accompanies it.
The central principle is the categorical imperative. Unlike a hypothetical imperative, which specifies what to do if a particular end is desired, a categorical imperative claims authority independently of such ends. In its formula of universal law, it directs an agent to act only on a maxim that the agent can at the same time will as universal law. A maxim is a practical principle specifying an action, its circumstances, and its purpose. The test therefore evaluates the principle of choice, not merely the physical description of a movement.
Kant’s example of a deceptive promise illustrates one kind of failure. Someone proposes to obtain money by promising repayment while intending not to repay. Universalizing this principle undermines the practice on which its success depends: assurances made under that principle could not perform the trust-generating role the deceiver needs. Kant distinguishes such contradictions in conception from contradictions in will. A general refusal ever to assist others may be conceivable, but he argues that a rational agent cannot consistently will it while also willing the assistance that the agent may need.
The formula of humanity directs agents to treat humanity, in themselves and others, always also as an end and never merely as a means. “Merely” is crucial. Ordinary cooperation involves people serving one another’s purposes, but it need not reduce anyone to an instrument. Deception and coercion are central cases because they bypass or subvert another person’s capacity to choose. This formulation links duty to moral autonomy: moral authority arises from rational self-legislation, not simply from externally imposed commands or the satisfaction of existing preferences.
Kant also distinguishes perfect from imperfect duties. Perfect duties impose relatively determinate constraints, whereas imperfect duties, such as beneficence and self-cultivation, allow latitude in how and when they are fulfilled. His ethics therefore includes positive obligations and the adoption of ends, not only prohibitions. Interpretation remains contested over how the formulations of the categorical imperative relate, how maxims should be specified, and how strict particular duties are. Kant’s 1797 essay denying a supposed right to lie from philanthropy is especially important because it exposes the tension between his stringent treatment of truthfulness and emergencies in which deception appears necessary to protect a threatened person.
Plural duties and contractualist justification
Not all deontological theories derive duties from one supreme principle. W. D. Ross argued that ordinary moral experience reveals several basic kinds of obligation. His account includes fidelity, reparation, gratitude, justice, beneficence, self-improvement, and non-maleficence. A promise creates a reason to keep it; an injury one has inflicted creates a reason to make amends; the possibility of benefiting another creates a reason to help. These reasons are not interchangeable expressions of a single requirement to maximize value.
Ross called these prima facie duties. The expression does not mean merely apparent duties that disappear upon closer inspection. It refers to genuine moral considerations that can be outweighed by other considerations in a particular situation. A promise to meet a friend may be overridden by the need to help someone seriously injured on the way. The promise nevertheless mattered: its moral significance was not an illusion produced by incomplete information.
An actual duty is what ought to be done after the relevant considerations have been assessed together. Ross denied that a mechanical formula could reliably determine the outcome in every case. Mature judgment and attention to circumstances are indispensable, and judgments about particular acts are less certain than recognition of basic duty-types. This position has affinities with ethical intuitionism, but it is not an endorsement of whatever immediately feels right. Reflection can expose mistaken descriptions, neglected obligations, and inconsistent priorities.
A different route to nonconsequentialist conclusions is contractualism, especially the account developed by T. M. Scanlon in What We Owe to Each Other (1998). Its central test asks whether an action would be disallowed by principles that no one could reasonably reject as a basis for informed, unforced general agreement. The test concerns justification to individuals rather than maximizing a total of benefits. A grave burden imposed on one person cannot automatically be justified by adding together very small benefits to many others. Reasonable rejection is not simply whatever someone happens to dislike; it depends on the strength and relevance of the reasons available to affected individuals.
Contractualism differs both from an actual contract and from every theory called contractualist. It does not require that people have historically consented to moral rules. Nor is it identical with John Rawls’s account of justice, which uses a hypothetical original position to select principles for the basic structure of society. Rawls’s A Theory of Justice (1971) nevertheless illustrates a related resistance to sacrificing the basic interests of some simply to increase an aggregate social total. Rossian pluralism and contractualism thus exemplify different foundations for deontological judgment: one emphasizes irreducible moral reasons, while the other emphasizes principles justifiable to each person.
Rights, personal constraints, and moral permissions
Many deontological theories explain restrictions on conduct through rights. A right can identify a protected interest or status that others must respect. The right not to be assaulted, for example, normally places others under a duty not to assault its holder. Rights are not all of one logical type: a claim against interference differs from a liberty to act, a power to alter a legal relationship, or an immunity against another person’s authority. Consequently, saying that a person “has a right” does not by itself specify every corresponding obligation.
Patient-centered theories explain constraints primarily by reference to the person acted upon. The objection to killing an uninvolved person to obtain organs for several patients is not exhausted by the resulting fear or damage to medical trust. On a rights-based account, the person has a claim against being killed and used in that way even in a carefully constructed case where wider consequences are excluded. The person’s status constrains how benefits may be pursued. Consent can sometimes alter these relations, but its power is limited: theories disagree over which protections can be waived and under what conditions.
Agent-centered theories instead emphasize the agent’s own actions, commitments, or relationships. A prohibition against committing a murder can give each agent a special reason not to murder, rather than simply a reason to minimize the total number of murders committed by anyone. These reasons are often described as agent-relative because their content depends essentially on the agent’s position. Special obligations to children, friends, or promisees are further examples. An agent-neutral reason, by contrast, recommends an objective without assigning it specifically to one person’s agency.
This distinction generates a characteristic puzzle. If committing one wrongful act would prevent several comparable wrongs by others, why should refusal be morally preferable? A deontologist may answer that responsibility is not reducible to managing the global total of wrongdoing. The distinction between respecting a person and promoting general respect for persons matters: violating someone’s rights in order to increase rights-respect elsewhere may contradict the very relationship the right establishes. Critics question whether this makes the agent’s moral integrity excessively central.
Moral permissions raise a related but separate issue. Some theories allow agents to give additional weight to their own projects and relationships without requiring them to do so. A person may permissibly spend time learning music rather than devoting every available hour to charitable work, even if the latter would produce greater benefits. Such permissions need not license indifference to severe need. They instead mark a space in which morality permits choices without demanding the impersonal best. Theories differ over the extent of this space and over whether it is best explained through autonomy, legitimate partiality, or limits on what others may demand.
Intention, doing harm, and the doctrine of double effect
Deontological reasoning often distinguishes what an agent intends from what the agent foresees. An intended harm is adopted as an end or as a means to an end; a foreseen harm is anticipated without necessarily forming part of the plan. The distinction is not equivalent to wishing harm versus regretting it. Someone may regret injuring an innocent person while nevertheless deliberately using that injury as the means to achieve a goal. Conversely, awareness that an action carries a harmful risk does not establish that the harm is intended.
The doctrine of double effect gives this distinction a specific role. Its historical background includes Aquinas’s discussion of self-defense in the Summa theologiae, although the familiar multi-condition formulations developed later. Versions commonly require that the act be permissible in its object, that the harmful effect not be intended either as an end or as a means, that the good effect not be achieved through the bad effect, and that there be a proportionately serious reason for permitting the harm. The doctrine is therefore not a general exemption for anything described as an unintended consequence.
The trolley problem, introduced into modern discussion by Philippa Foot in 1967 and developed by Judith Jarvis Thomson, tests these distinctions. In a standard switch case, redirecting a runaway trolley saves five people but kills one person on another track. In a footbridge case, pushing a person into its path stops it and saves five. The numerical outcomes can be held constant while the victim’s causal role differs. In the latter case, the person’s body is used as the means of stopping the trolley; in a standard interpretation of the switch case, the one person’s death is not needed for the rescue to succeed.
A separate distinction concerns doing harm and allowing harm. Not initiating a lethal process may differ morally from initiating it, but the distinction cannot simply be equated with bodily movement and inactivity. A person can allow harm through a deliberate action, and an omission can violate a stringent duty. A caregiver who withholds something indispensable from a dependent person may bear serious responsibility despite performing no direct assault. Existing duties, control over the situation, and the agent’s relationship to the endangered person all affect the analysis.
These cases also show why causation, intention, and permissibility must be examined separately. Causing a death does not by itself establish wrongful killing, while lacking an intention to kill does not establish permissibility. Negligence, recklessness, disproportionate risk, or a breach of protective duty can make an unintended injury wrong. Critics of double-effect reasoning further argue that descriptions of intentions can be manipulated. Defenders respond that the relevant plan must be identified through its practical and causal structure, not through an agent’s preferred verbal description.
Conflicting duties, catastrophe, and theoretical objections
Duty conflict is a central problem because deontological requirements can pull in different directions. Confidentiality can conflict with protecting someone from serious harm; fidelity to a promise can conflict with urgent assistance. Some theories deny that fully specified genuine duties can conflict, treating apparent conflicts as evidence that a rule or circumstance has been inadequately understood. Others accept conflicts among pro tanto duties and distinguish them from the final duty that governs the situation. A stronger position allows genuine moral dilemmas in which some requirement remains unfulfilled whatever the agent does.
Residual obligations help explain what can remain after a justified choice. Someone who permissibly breaks a promise to respond to an emergency may still owe an explanation, apology, or compensation. This does not necessarily show that the emergency response was wrong. It shows that the overridden claim retained moral significance. Such cases challenge accounts that treat the losing consideration as entirely erased once the best action has been identified, although consequentialist theories can also explain practices of apology and repair through their effects.
Catastrophic cases press the question of how strong constraints should be. Threshold deontology holds that a prohibition can bind against ordinary gains yet be overridden when the consequences become sufficiently grave. This avoids some implications of absolute prohibitions, but it creates difficult questions about where the threshold lies. If a slight increase in expected harm suddenly makes a previously forbidden act permissible, the transition may appear arbitrary. If harms accumulate across many cases, it is also unclear whether each act should be evaluated separately or as part of a larger pattern.
The so-called paradox of deontology asks why a theory should forbid an agent from committing one violation to prevent several similar violations. Closely related is the accusation that deontology values keeping one’s own conduct untainted over protecting potential victims. Replies distinguish constraints from impersonal goals: a right is not simply an instruction to maximize the number of occasions on which that right is respected. Some defenses emphasize the separateness of persons or limits on the authority any person has to impose sacrifices on another. Whether these explanations justify stringent constraints remains disputed.
A further debate concerns whether deontology can be redescribed as consequentialism by constructing a sufficiently elaborate ranking of outcomes. If outcomes include who acted, what was intended, and whose rights were violated, a ranking can reproduce many deontological verdicts. Critics argue that this maneuver may obscure rather than explain the difference between theories: reproducing a set of choices does not establish that the reasons behind those choices are consequentialist. Deontology faces an analogous explanatory demand. Listing prohibitions is insufficient unless the theory explains why those particular constraints bind, how their scope is determined, and how agents should reason when evidence is uncertain.
Professional practice, law, and emerging technologies
In professional ethics, deontological ideas often appear as duties attached to roles: confidentiality, truthfulness, loyalty, competent performance, and avoidance of exploitation. These duties can give professionals reasons that are not exhausted by maximizing immediate benefits. A lawyer’s obligation to a client or a researcher’s obligation to a participant may remain significant even when disclosure or deception would serve a useful purpose. Role duties nevertheless require justification; a professional code cannot make an otherwise wrongful practice permissible merely by declaring it part of the role.
In medicine and research ethics, informed consent illustrates the importance of respecting a person’s authority over participation and intervention. Beneficial expected outcomes do not by themselves authorize enrollment in an experiment without adequate permission. The Declaration of Helsinki, first adopted in 1964, embodies protections for research participants alongside concern for scientific and social value. These frameworks combine several kinds of ethical reasoning rather than expressing a single pure theory. Questions about capacity, voluntariness, disclosure, and emergency exceptions show that respecting autonomy requires more than obtaining a signature.
In law and political philosophy, rights-based constraints help explain why institutions may not treat every interest as exchangeable for aggregate welfare. Due process and protections against arbitrary punishment restrict how public purposes may be pursued. However, legal systems also employ proportionality tests, balancing, and emergency powers. Constitutional rights therefore should not be assumed to function as absolute moral prohibitions. A legal right’s interpretation, institutional enforcement, and permissible limitations are distinct questions from the philosophical justification of rights themselves.
Digital systems bring similar issues into new settings. Data privacy raises questions about whether beneficial prediction or efficient service provision can justify collecting and using information without adequate authorization. In artificial intelligence, deontological approaches motivate explicit constraints on system behavior, such as prohibitions on unauthorized disclosure or discriminatory treatment. Encoding a rule is not equivalent to solving its ethical interpretation. Categories such as consent, coercion, relevant difference, and foreseeable harm depend on context, and multiple requirements may conflict.
These applications expose an important distinction between a moral theory and a decision procedure. A theory may explain why deception is wrong without providing a computational test for every misleading communication. Conversely, a practical procedure can combine rules, risk assessment, human review, and institutional accountability without settling the ultimate foundation of morality. Deontological implementation therefore requires specifications of authority, exceptions, evidence, and responsibility. Uncertainty about whether an act violates a right cannot simply be replaced by certainty that a rule has been mechanically followed.