Doing and Allowing

The Moral Problem of Doing and Allowing

The distinction between doing harm and allowing harm concerns how an agent stands in relation to a harmful outcome, and whether the kind of involvement changes the justification required for the agent's conduct. Its most familiar manifestation concerns killing and letting die, but its structure extends to injury, deprivation, rights infringement, institutional omission, and the management of foreseeable risks. A person can cause a death through an intervention; a person can also knowingly decline to prevent a death that would otherwise occur. Both can be morally responsible. The philosophical question is whether the first relation normally carries a more stringent prohibition than the second and, if so, why.

The distinction operates at several levels that must remain analytically separable. One question is descriptive: what makes an agent count as doing harm rather than allowing it? A second is normative: does that difference affect what one is permitted or required to do? A third is evaluative: how does it affect blame, the severity of wrongdoing, or the agent's character? A fourth is institutional: which duties and liabilities should law, medicine, politics, and organized practices attach to interventions and omissions? To collapse these questions is to make the ethical answer depend upon an arbitrary choice of terminology.

In ordinary moral judgment, refraining from assault seems more demanding than sacrificing every resource that might prevent someone else's injury. Yet parents owe affirmative protection to children, clinicians acquire duties toward patients, and public officials may incur obligations whose breach is grave precisely because they failed to act. The central philosophical task is therefore to identify the relationship among harm production, protective responsibility, justified liberty, and the moral standing of those exposed to harm.

This study defends a qualified priority of nonimposition: when an agent introduces a harmful threat or appropriates another person's protected sphere to secure a benefit, that act ordinarily requires a stronger justification than a failure to prevent an independently arising threat. The qualification matters. Moral responsibilities already undertaken, dependence, ownership of an earlier danger, control of institutions, and the costs of rescue can convert an omission into a serious wrong and can sometimes defeat permissions to refrain. The argument will reconstruct rather than assume the conceptual distinction and will examine objections before presenting its conclusions.

Definition, Scope, and Logical Architecture

Doing harm may provisionally be defined as contributing to a harmful outcome through a relevant positive element of one's conduct in the process that produces that outcome. Allowing harm may provisionally be defined as failing to interrupt or prevent an independently developing harmful process, when one had a relevant opportunity to intervene. These are working characterizations, not an assertion that every disputed example has a uniquely correct ordinary-language description. Their exact conditions are the substance of competing causal, counterfactual, process, and rights-based analyses.

The objects of classification must be specified. One must identify an agent, an outcome or threatened interest, a temporally bounded episode of conduct, the relevant alternative possibilities, and a proposed explanation of how the outcome occurred. If a physician stops an intervention, the description 'withdrawing treatment' tells us what happened physically. It leaves open whether the physician killed, allowed a disease to kill, ceased providing aid, or wrongfully deprived someone of an entitlement. Descriptions can be true simultaneously at different levels. A theory must explain which description carries the morally significant structure.

Harm must also be distinguished from a mere failure to benefit. Suppose an individual already faces a lethal danger. A passerby has an opportunity to prevent death but declines. Relative to the counterfactual in which rescue occurs, the person is worse off; relative to the history without rescue, the danger develops independently. Neither counterfactual alone defines the normative baseline. Rights, prior commitments, legitimate expectations, and the origin of the threat determine whether a loss should count as an infringement, a failure to provide a benefit, or both.

The Doctrine of Doing and Allowing is best expressed as a principle about differential justification. Other relevant factors being suitably controlled, doing harm generally requires stronger reasons than merely allowing harm. The principle neither asserts that every act of harm is impermissible nor licenses every failure to help. A defensible account must state the comparison class, explain how it interacts with rights and duties, and allow exceptionally strong rescue obligations. Quinn (1989), Scheffler (2004), and Woollard (2015) provide three important but different defenses of this structure.

Historical Antecedents and the Formation of the Debate

Ancient and medieval philosophy supplies important conceptual antecedents without furnishing a modern theory in its contemporary form. In Nicomachean Ethics III.1–5, Aristotle examines voluntariness, force, ignorance, and the assignment of praise and blame. His discussion helps explain why conduct cannot be assessed apart from the agent's knowledge, capacities, and circumstances. Aristotle's distinction between voluntary and involuntary action should not be retroactively equated with doing and allowing. It clarifies a connected question: under what conditions is conduct attributable to an agent?

Aquinas's discussion of self-defense in Summa Theologiae II–II, question 64, article 7, introduces a significant analysis of one action having different effects and of the role of intention and proportionality. Later discussions of double effect draw upon this text. It does not supply an exhaustive medieval analysis of omissions, nor does the intention/side-effect distinction map exactly onto the doing/allowing distinction. An intentionally permitted death and an unintentionally inflicted death are both conceptually possible.

In the development of modern moral and political reasoning, the contrast between duties of noninterference and duties of aid was increasingly articulated through rights and welfare. The modern debate acquired distinctive shape when advances in medical practice, the ethical evaluation of life-sustaining treatment, humanitarian responsibilities, and objections to unrestricted consequentialism made vivid cases unavoidable. The twentieth-century literature brought separate questions about agency, causation, rights, and outcomes into direct dispute.

Philippa Foot's 1967 essay supplied a decisive turning point by placing the distinction between what agents do and what they allow alongside the doctrine of double effect. James Rachels (1975) challenged an apparent moral asymmetry between active and passive euthanasia. Jonathan Bennett's analyses, Warren Quinn's 1989 defense, Judith Jarvis Thomson's trolley cases, and subsequent work by Kagan, Kamm, McMahan, Scheffler, and Woollard transformed familiar moral judgments into a sustained investigation of how harmful outcomes are related to agency.

The history is therefore best understood as the convergence of several traditions of inquiry: classical responsibility, negative and positive duties, causal explanation, medical ethics, and modern deontological theory. Their convergence explains the durability of the problem, while their distinct origins explain why an answer confined to the vocabulary of 'acts' and 'omissions' remains inadequate.

Foot: Rights, Interference, and Harmful Processes

In 'The Problem of Abortion and the Doctrine of the Double Effect' (1967; reprinted 1978), Philippa Foot examines cases in which a person must decide whether harm to one may be brought about to prevent harm to others. Her famous judge and trolley examples reveal that outcome counts alone fail to determine moral assessment. The judge who executes an innocent person to prevent a riot commits an intervention against that person. The trolley driver who turns a vehicle away from five toward one faces a different configuration of threats and responsibilities. The examples were introduced to investigate the relationship of intended harm, foreseen harm, and duties owed to individuals.

Foot's approach associates the prohibition against harmful interference with negative rights and the duty to provide help with positive rights. Where people possess claims against being attacked or interfered with, an agent cannot readily trade those claims for the aggregate advantage of others. Duties of assistance may be serious, but the form and strength of the claim differ. Foot's later work develops an account of initiating or entering sequences that lead to harm and indicates that doing and allowing cannot simply be translated into bodily movement and passivity.

A complication immediately appears. If a person who has contracted to perform at a theater deliberately stays away, that absence may spoil a performance: an omission is explanatorily important to a harmful outcome. Conversely, arranging the cancellation of financial support can be an active way of ceasing to provide aid. Thus Foot's insight requires a theory of what counts as 'interference' and what makes a causal sequence independent, rather than the assertion that hands moving always equal doing and hands still always equal allowing.

Her central contribution is normative as much as causal. The difference between intruding upon a person's protected position and withholding an optional benefit can ground distinct requirements. Whether this grounds all apparently permissible trolley diversions becomes doubtful once a bystander, rather than a driver already piloting the vehicle, makes the choice. That challenge is developed by Thomson and contemporary trolley literature.

Rachels and the Challenge of Moral Equivalence

James Rachels's 'Active and Passive Euthanasia' (1975) is a landmark critique of a categorical preference for letting die over killing. He considers two malicious agents, conventionally called Smith and Jones. Smith actively drowns a child; Jones plans to do so but finds the child already in danger and deliberately withholds rescue. Both welcome the same death for the same self-interested reason. The moral reaction that their conduct is comparably reprehensible is intended to undermine the assertion that killing is, simply as killing, worse than letting die.

The argument has a clear structure. First, there are pairs of cases apparently matched with respect to motive, intended outcome, and preventability, while differing in intervention versus nonintervention. Second, if the distinction itself always generated a major difference in moral wrongness, the two cases would differ correspondingly. Third, they appear morally equivalent. Therefore, the difference alone cannot explain the moral condemnation. Applied to medicine, the argument challenges treating the mere absence of a new lethal intervention as a complete justification for an avoidably prolonged or painful death.

A serious qualification concerns the inferred conclusion. From the equal blameworthiness of two egregious wrongdoers it does not follow that the justificatory threshold for every active harm and every omission is equal. Two agents may violate different rights with equal malice. A normative distinction can govern what would justify conduct without producing a fixed additive increment of blame in every possible comparison. Kagan's criticism of the assumption that each morally relevant factor must produce a measurable difference in each pair of otherwise matched cases illuminates this point.

Rachels also shows something supporters of the doctrine must accept: the presence of an omission can never function as a moral absolution. Jones deliberately exploits a vulnerable child's danger; his desire, opportunity, and possibly specific responsibility matter independently of his failure to set the lethal process in motion. The lasting contribution of the contrast is to demand a theory robust enough to judge omissions seriously while explaining, if possible, why some interventions remain specially constrained.

Quinn: Positive and Negative Agency

Warren Quinn's 'Actions, Intentions, and Consequences: The Doctrine of Doing and Allowing' (1989) formulates a defense built upon the contrast between positive and negative agency. What matters is the kind of contribution through which the agent becomes relevant to the harmful outcome. Positive agency typically involves action that directly contributes to the outcome; negative agency concerns a failure to prevent it. His proposal asks what most directly explains the upshot rather than merely whether the agent moved.

The moral defense rests on persons having stronger claims against harmful intrusion than against every failure of rescue. The connection is not simply that active agency mechanically produces more evil. Rights create different thresholds of permissible justification. Quinn accordingly distinguishes the stringency of a prohibition from how blameworthy someone becomes after violating it. A malicious failure to rescue can be as deeply blameworthy as a malicious killing even if their justificatory structures differ.

The structure can be reconstructed as follows. Individuals require effective claims protecting their bodies and interests against becoming resources for others. Such claims would be radically weakened if agents could directly impose major injuries whenever doing so delivered a somewhat larger benefit elsewhere. Moral practice therefore recognizes a stronger barrier to positive harmful imposition than to an omission in the face of an independently occurring danger. This barrier survives the possibility that some failures to save are impermissible under more specific obligations.

The objection concerns whether the agency distinction genuinely explains the rights or redescribes them. If rights are already needed to determine whether an interruption is wrongful, the definition of 'doing harm' may become morally circular. Moreover, sequences can include prior acts, mechanisms, or omissions that blur the directness of contribution. Quinn's achievement is a sophisticated justification-based account; its incompleteness lies in the need for more precise criteria concerning complex causal pathways and ownership of previous threats.

Intention, Double Effect, and Their Separate Roles

The doctrine of double effect distinguishes the moral significance of intended harm from harm foreseen as a side effect, under conditions involving the character of the act, the importance of the good pursued, and proportionality. Quinn explored that doctrine in a separate 1989 article in Philosophy & Public Affairs. The distinction is analytically independent of doing and allowing. One can intend a death by declining aid; one can bring about an unintended death through a dangerous rescue; one can foresee harm and nevertheless be directly involved in its production.

A typical example is a clinician administering treatment that relieves otherwise unbearable suffering while foreseeing but not aiming at a potentially life-shortening consequence. The interpretation of such cases depends on medical facts, dosage, causal mechanism, available alternatives, and informed consent. Neither a label of intention nor an active/passive description resolves their justification. The foundational inquiry asks what was done to whom and how the intervention relates to their protected interests.

The differences matter because the grammar of 'letting' sometimes conceals intention. Someone can deliberately let a dependent person die to obtain money. Likewise, 'killing' does not guarantee a homicidal intention: a driver can foresee a death as a tragic unavoidable side effect of steering away from a greater collision. A systematic moral judgment must use distinct variables for causal role, intended end, intended means, foreseen side effects, and the agent's authority to expose others to risk.

The two doctrines can support one another where intentionally using another person as a means also involves introducing a harmful process. They may diverge when rescue generates an unintended harm or when withholding aid is itself a deliberate strategy to bring death about. Philosophically, the most informative theory allows intersection without declaring the distinctions synonymous.

The Trolley Problem and the Limits of Simple Classification

The trolley family of cases tests whether killing one person to save several others is permissible when the harm falls upon an uninvolved individual. Foot's original driver case allowed a reading in which the driver would actively kill whoever lay in the route of the vehicle; diverting it reduces deaths. Thomson's bystander variant removes the special relation of being the driver: the bystander can intervene to switch a runaway trolley toward one person rather than allow it to continue toward five.

The apparent permissibility of diversion challenges a simple prohibition on doing harm when allowing a greater harm is the alternative. Philosophers have proposed further distinctions involving redirecting a preexisting threat, using a person as a means, ownership of the risk, and the degree of agency involved in producing a new threat. Each proposed explanation must survive altered variants: a trolley redirected from five to five, a pedestrian pushed onto the tracks, a barrier that would block the vehicle, or a switch that merely transfers the threat to another location.

Thomson's 'The Trolley Problem' (1985) made the bystander formulation prominent. In 'Turning the Trolley' (2008), she reconsidered permissibility and challenged the assumption that an agent may impose on a stranger a burden the agent would refuse to bear herself. This is a historical development within one author's argument, not a settled consensus. The revision illustrates why trolley intuitions are philosophical data in need of interpretation rather than an authority determining moral theory.

The dossier's conclusion is modestly determinate. The doing/allowing distinction alone does not solve all trolley cases. In some configurations the origin and redirection of the threat explain why a direct intervention may be permissible, but a defender must also consider whether the victim becomes an instrument and which claims are overridden. Cases designed to isolate one variable can illuminate that variable while producing unrealistic assumptions about time, knowledge, and alternatives. Those limitations become especially important when trolley analogies are imported into engineering.

The Causal Analysis and Its Difficulties

An intuitive causal analysis says that an agent does harm by causing it and allows harm by failing to prevent it. In the simplest case, a person releases a dangerous object toward a victim; another sees a naturally falling object and chooses not to stop it. The first is causally implicated in starting the dangerous process; the second is located at a potential point of interruption. A metaphysical problem arises because omissions can themselves figure in true causal explanations. We say a plant died because its caretaker neglected to water it, or a patient deteriorated because essential care was withheld.

The dispute cannot be resolved by assuming all genuine causation requires positive physical transfer. Causal explanations are contrast-sensitive: 'because the door was not closed' may explain the entry of smoke given a background condition that the door could have been closed. Yet the explanatory relevance of a missing prevention does not automatically make the omission the same kind of contribution as opening a gas valve. The challenge is to analyze the difference without declaring causal responsibility either present everywhere or absent in every omission.

Three cases sharpen the point. First, a person fails to supply a necessary protective shield, allowing an existing threat to reach someone. Second, the person removes a shield already in place. Third, the person introduces the threat itself. Causal dependence may connect the resulting harm to the agent in all three, while their normative character can vary with who owned the shield, who installed it, and who held a duty to maintain it.

A defensible causal account therefore needs an account of relevant processes and the baseline against which contribution is classified. Causation does essential explanatory work, but causal connection alone does not establish the structure of the moral claim. The question 'What caused this injury?' and the question 'Who was entitled to make this contribution?' are logically related and conceptually distinct.

Counterfactual Dependence and the Problem of Baselines

Counterfactual analysis asks whether harm would have occurred if an agent had behaved differently. This is indispensable for identifying opportunities and responsibility. It is nevertheless insufficient as a definition of allowing, because nearly every preventable harm would disappear under some rescue action. The counterfactual 'Had I helped, the person would have survived' identifies preventability, while leaving unchanged the question whether my contribution is an intervention or a failure of rescue.

Likewise, a single harmful action can be overdetermined. Two independent toxins may have been administered, each sufficient for death. If either dose were removed, the victim might still die. A naive but-for test denies the causal relevance of both, although each agent introduced a lethal process. More sophisticated accounts examine actual sequences, causal sufficiency, and the relations among contributing conditions. They must also consider uncertainty about background facts and relevant possible worlds.

A moral theory needs a normative baseline in addition to a descriptive counterfactual. If an agent restores a person to a condition the agent had previously wrongfully imposed, that restoration may be owed rather than charitable. A passerby who gives up an optional benefit occupies a different position from a trustee who fails to transfer property already belonging to someone else. The material consequences can match while the entitlement structure changes the moral description.

The useful counterfactual question is thus indexed: against which legitimate background of rights, commitments, existing dangers, and available alternatives is the decision assessed? It is a mistake to choose whichever baseline makes a desired verdict trivially true. Independent specification of entitlements and the provenance of danger limits this arbitrariness. The need for such specification foreshadows why nonimposition, prior responsibility, and rescue form a connected but nonidentical moral system.

Bennett and the Range of Alternative Behavior

Jonathan Bennett investigates positive and negative relevance with unusual attention to the logic of causal explanation and the information supplied by facts about conduct. His discussions in 'Morality and Consequences' and The Act Itself challenge the idea that ordinary verbs such as 'make', 'let', or 'allow' can provide a complete philosophical criterion. Descriptions of an agent can be more or less informative; an explanation may cite a highly specific action where a less specific omission would already explain the outcome.

On a simplified formulation attributed to Bennett, conduct is positively relevant where most alternative ways the agent could have behaved would not have produced the harmful outcome, and negatively relevant where most alternatives would have left the harmful outcome in place. Spraying poison on a victim is positively relevant because most of the things the agent could instead have done would have avoided poisoning. Failing to pull someone from water is negatively relevant because most possible behaviors would not have supplied rescue.

The significance lies in separating the difference in involvement from the superficial distinction between bodily movement and stillness. A person might vigorously run away rather than rescue someone; physical activity does not transform the omission into producing the drowning. A person may fail to turn up for work, thereby directly derailing an event that depends on the person's attendance; passivity does not settle all causal descriptions.

The conceptual difficulty is the measure over possible conduct. There is no neutral way to count 'most' alternatives without specifying a space of possible actions, levels of description, feasible capacities, and time scale. An absurdly large number of trivial motions may overwhelm the relevant options, and the most salient alternative may matter more than the number of imaginable alternatives. Bennett's account therefore reveals an important structural insight but needs further argument before it can be treated as a determinate moral criterion.

Process Accounts, Threats, and Substantial Facts

Process accounts focus on whether the agent's conduct enters the sequence leading to harm. The account avoids equating allowing with an inactive body. A person can allow harm while performing an unrelated action; conversely, a failure to perform a role can become an essential part of a harmful process. Foot's emphasis on sequences and later elaborations by Fiona Woollard connect the metaphysical classification to the difference between a dangerous process and an opportunity to interrupt it.

Woollard (2015) proposes, at the level developed in her published chapter descriptions, that facts about an agent's behavior count as part of the sequence leading to harm when they are substantial facts rather than merely background or non-substantial conditions. The distinction is designed to handle complicated cases without letting every nonperformance count as an active harmful step. It remains vulnerable to disputes over which facts qualify as substantial, especially when a social role establishes what a person was expected to do.

A physical sequence often has several layers: the release of a threat, enabling conditions, protective barriers, and possible interventions. A person who knowingly opens a barrier in the path of an approaching hazard may seem to do more than someone who simply leaves the barrier unbuilt. Yet an owner who discontinues access to a privately maintained protective measure may have a different moral claim from an attacker who removes another person's safety mechanism. Process analysis must be supplemented with entitlements and the history of provision.

The important synthesis is that harmful imposition has both a causal and normative dimension. The causal question locates how conduct affects the route by which harm arrives; the normative question determines whether the agent has intruded upon a sphere another person has standing to protect. Process accounts make the first question rigorous and prevent the second from disappearing behind the word 'inaction'.

Kamm and Context-Sensitive Moral Significance

Frances Kamm develops a detailed case-based method in which moral significance depends on interactions among an agent's conduct, the victim's entitlements, the alternative available to the agent, and the way help would be supplied. Her discussions in Morality, Mortality and Intricate Ethics contest the assumption that each moral feature has an invariant, additive effect across all possible contexts. An act/omission difference can be real even where a pair of malicious agents is judged similarly, and the difference can matter through the structure of justification rather than through a simple ranking of wickedness.

Kamm emphasizes that an individual may have a stronger claim to what is already independently theirs than to assistance another person would have to provide at significant cost. In some letting-die cases the person loses an opportunity that depended on the agent's fresh help. In some killings, by contrast, the agent actively removes the person's independently secured prospect of continued life. The distinction can be morally relevant because it tracks whose protected position is being altered.

Her analysis also complicates any mechanical mapping from killing to strong prohibition. Particular killings may possess features structurally closer to some omissions, and some cases of withdrawn protection resemble interference. The proper question is which person would have to surrender which resource, body, liberty, or existing protection, and why that surrender is or is not within another's authority. This shifts attention from action labels toward an account of justified burden shifting.

A critical concern is whether multiplying finely specified case distinctions becomes too dependent upon disputed intuitions. If one can always add a new principle to fit a counterexample, the theory risks becoming unfalsifiable within its own method. Nevertheless, Kamm's insistence on contextual interaction is a necessary corrective to the inference from one pair of morally equivalent cases to an unrestricted moral symmetry thesis.

Scheffler: Responsibility and the Point of Agency

Samuel Scheffler (2004) argues that the role of doing and allowing arises from how persons understand themselves as agents answerable to moral requirements. A moral requirement is addressed to someone capable of regulating conduct for reasons. Such regulation assigns particular responsibility to what the agent does, and that practical orientation supplies a basis for taking a distinction like doing/allowing seriously. The proposed defense addresses the ethical significance of agency at a level deeper than a definition of bodily action.

One version of the argument can be reconstructed as follows. To deliberate under moral standards is to regard oneself as accountable for conduct. Accountability requires perceiving the regulation of one's conduct as a central task of agency. Therefore, some special responsibility for what one does must enter the normative structure that agents inhabit. A morality that treats every action solely as an instrument for producing the most desirable aggregate outcome risks missing the distinctive standpoint from which agents regulate their own behavior.

The main objection concerns the step from psychological and practical self-understanding to a stronger moral prohibition. We are agents who can also deliberately decide not to rescue; omissions are often objects of reflection and choice. Seeing oneself as especially responsible for one's acts may reflect an understandable but unjustified bias. Bradley and Stocker (2005) criticize the attempt to derive the normative distinction from the structure of agency, raising questions about what the premises establish.

Scheffler's argument has enduring value even if it is insufficient as a stand-alone proof. It explains why responsibility for self-directed interventions may be morally salient without implying that omissions disappear from the domain of agency. The reconstructed conclusion is a constraint on treating persons solely as organizers of outcomes; supplementary argument is needed to determine when this constraint licenses declining rescue and when institutional responsibility requires affirmative conduct.

Woollard: Nonimposition and the Meaning of Belonging

Fiona Woollard's Doing and Allowing Harm (2015) offers one of the most sustained defenses of the doctrine. It links the distinction to protection against the imposition of harmful burdens. For a body, a resource, or another protected interest genuinely to belong to a person, morality must place substantial restrictions on how others may direct harmful processes toward it. It must also acknowledge some liberty to decline the use of one's body and resources for the rescue of others. Both restrictions and liberties support coherent personal boundaries.

The argument begins from the significance of independent entitlement. If another person could compel the use of my body whenever doing so achieved a greater good, self-ownership would have little practical content. If I could impose a physical harm on another merely to preserve a slightly greater advantage for myself, that person's protected sphere would be equally insecure. A moral framework can therefore impose stringent constraints on harming while retaining qualified permissions not to prevent every harm.

Woollard's work also treats this permission as limited. A person who encounters a child drowning in shallow water may have an affirmative duty to help when the costs are minor and the danger grave. The permissibility of refraining from rescue cannot be derived directly from the classificatory fact that a death would be allowed. The justification must consider proportional costs, the significance of rescue, and the duties under which the agent acts.

The framework invites two further questions. First, what exactly makes a burden count as an imposition rather than an unfulfilled request for assistance? Second, how can the theory recognize genuine ownership and bodily protection in a world where resources and institutional capacities are historically dependent on collective action? Answering these questions requires connecting local cases to legal rights, relational duties, and systemic harms while preserving the core insight that protected belonging has moral consequences.

Safety Nets, Withdrawal, and Denial of Resources

Safety-net cases present the sharpest challenge to a simple doing/allowing division. An agent may remove a barrier that had been preventing an independently approaching threat; disconnect equipment providing life support; terminate funding that has kept a vulnerable person alive; or withdraw someone from a program on which continued safety depends. The immediate event is an action. The harmful trajectory may have originated elsewhere. The normative verdict depends heavily on whether the protection was owned, promised, required, or voluntarily supplied.

McMahan's 'Killing, Letting Die, and Withdrawing Aid' (1993) investigates the criteria for classifying these circumstances. The act of withdrawal cannot automatically be assimilated either to lethal aggression or to pure nonassistance. It may restore a prior threat to operation, interfere with another person's protection, or fail to continue a benefit that the provider was entitled to discontinue. The same physical procedure can have a different moral significance when the claimant's rights and the provider's duties differ.

An important refinement appears in analyses that distinguish enabling harm from doing harm and merely allowing it. Barry and Øverland (2016) attend to cases in which an action makes harm possible without directly introducing a complete harmful process. The resulting category illuminates situations such as removing protection or denying access to resources. It may support intermediate justificatory demands, though its precise boundaries are contested.

A persuasive account should avoid an incentive to withhold initial aid so as never to become bound by it. If rescue automatically creates a perpetual obligation greater than the original duty, agents may become reluctant to provide temporary help. Yet the opposite rule—every aid relationship may be abandoned without regard to reliance—undermines justifiable expectations. The solution must incorporate reasonable notice, transfer of responsibility, alternatives, agreements, and the vulnerability produced by induced reliance.

Letting Oneself Do Harm and the Temporal Problem

Cases involving one's own earlier wrongdoing challenge an exclusive focus on the action currently performed. Suppose an agent secretly poisons a drink, regrets nothing, and later could prevent the intended victim from consuming it by giving a warning. If the agent remains silent, the immediate conduct resembles an omission. Yet the death would complete a harmful process the same agent already initiated. Calling the present failure mere allowing appears to discount the moral relationship established by the earlier act.

The problem has been developed by Persson (2013) and Hanna (2015), as analyzed in the current Stanford Encyclopedia discussion. Their cases distinguish allowing oneself to have done harm from allowing an independently existing danger. In the poison example, the agent's corrective duty arises through authorship of the threat. The warning would not be a discretionary favor to a stranger. It would be an urgently required intervention to prevent one's previous conduct from taking effect.

A related case concerns an involuntary spasm that will discharge a weapon unless restrained. Does failure to restrain oneself constitute killing, allowing harm, or allowing one's body to produce harm? The agent's foresight and capacity for control matter, but the moral question is also whether the dangerous mechanism is part of the agent's sphere of responsibility. Time-sensitive classifications need to represent continuous agency rather than isolating single instants.

The larger lesson is that a person cannot erase responsibility for an initiated risk by moving into a later posture of inaction. An adequate doctrine includes temporal continuity: the originator of a threat may have a special duty to prevent its completion, even where a bystander would not be similarly obligated. This condition does not dissolve the doing/allowing distinction; it shows how previous doing can alter what later allowing means.

Omissions, Knowledge, Capacity, and Special Duties

An omission becomes morally assessable where an agent could reasonably have acted and the neglected act bears a relevant relation to what was owed. Knowledge, opportunity, ability, the cost of intervention, and the presence of a duty determine the analysis. A child left unfed by a responsible caregiver is not merely an unfortunate victim whom someone declined to help. The caregiving relationship establishes a positive obligation, and the omission violates it directly.

The category of special duties includes promises, contractual commitments, professional responsibilities, entrusted authority, voluntarily assumed care, and responsibility for dangers one has produced. These duties can vary in strength and scope. A lifeguard on duty, an ordinary passerby, and the person who caused a swimmer to enter peril occupy different moral positions, even if the rescue task is physically identical. To say all three 'let the person drown' supplies only an incomplete description of their obligations.

Capacity matters because demands that exceed what an agent could reasonably do are not ordinary violations of duties of rescue. Uncertainty complicates this principle: the agent may have been obliged to investigate danger, seek assistance, or adopt a precaution despite incomplete knowledge. Deliberate ignorance can itself be culpable where a role includes an obligation to remain informed. The resulting responsibility can be direct even if the eventual harm comes through a natural or independent process.

The distinctive problem of omissions thus lies partly in identifying which action was required. There are infinitely many things a person did not do. A defensible attribution selects a relevant counterfactual action because the agent occupied a position of duty, capacity, or special connection, not merely because the harmful outcome can be imagined away by altering behavior. This guards against the inflation of responsibility while allowing serious condemnation of neglected obligations.

Rights, Bodily Integrity, and Positive Duties

Rights-based approaches offer a normative explanation of why introducing harm differs from failing to prevent harm. A right against assault protects bodily security from a class of intrusions. A right to assistance, when recognized, places another person or institution under an affirmative duty to provide support. Their forms differ because the protected activity, claimant, duty-bearer, costs, and background social arrangements differ. The assertion that negative rights are generally stronger requires argument and does not entail that every negative right overrides every positive right.

The standard bodily example concerns the use of one person's organs to save several others. Even if organ extraction would maximize survival, the protected person's authority over their own body normally blocks treating them as available medical material. By contrast, a person may have an obligation to donate money or render low-cost aid when another is at grave risk. The difference is supported by the moral importance of bodily integrity and the way imposition would make others' bodies instruments for general welfare.

But property, money, and social resources are not always analogous to bodily organs. Legitimate taxation, emergency services, and public health measures presuppose enforceable positive responsibilities. The institutional allocation of duties can alter what a person is entitled to keep or withhold. A patient legally and morally entitled to necessary care may be harmed through an institutional refusal, even when no individual administrator performs the bodily injury.

A plausible rights theory therefore identifies rights and correlative duties independently of the doing/allowing classification. Otherwise it risks circularity: it declares harmful conduct to be 'doing' when it violates a negative right and then treats the violation as more serious because it is doing. Once independent entitlements are made clear, the distinction has explanatory force as one feature of why some claims against imposition and some claims to rescue are structured differently.

Consequentialism, Agent-Relative Constraints, and Demandingness

Consequentialist approaches assess available courses of conduct by reference to the value of their consequences, commonly without a foundational preference for bringing about or failing to prevent an outcome. If preventing five deaths by harming one person makes the overall outcome better, a simple maximizing principle favors intervention. Likewise, if a small sacrifice can avert a serious preventable death, the agent has strong reason to help irrespective of proximity. Such approaches expose the moral costs of treating nonintervention as automatically protected.

Deontological approaches grant independent force to rights, restrictions, and the moral status of means. They may prohibit certain impositions even when those impositions would reduce total harm. The dispute is not captured by asserting that one side cares about outcomes and the other ignores them. A credible nonconsequentialist theory must consider the gravity of what is at stake, and credible consequentialist theories may recognize the large instrumental benefits of stable rights and publicly predictable rules.

Kagan's The Limits of Morality develops an influential attack on both restrictions against causing harm for the greater good and permissive limits on required beneficence. Singer's 'Famine, Affluence, and Morality' (1972) emphasizes the duty to prevent serious suffering when prevention requires no comparable sacrifice. The familiar shallow-pond rescue sharpens the challenge: if one must save a nearby child at trivial cost, distance and the existence of many other potential rescuers require independent justification before they reduce moral concern.

A defensible theory should answer two opposing pressures. Unlimited demands to maximize welfare can sacrifice protected personal projects and control over the body; unlimited licenses to refrain from rescue can tolerate grave avoidable suffering. The best interpretation of doing/allowing treats the doctrine as a pro tanto constraint against harmful imposition while recognizing affirmative duties proportionate to cost, role, and urgency. It thereby rejects both categorical exemption and an automatic numerical calculus.

Contractualism and Reciprocal Justification

Contractualist analysis asks which principles governing conduct could be justified to individuals who must live under them. For doing and allowing, this changes the argumentative form. The question is whether persons have reason to reject a principle permitting others to injure them whenever doing so would deliver somewhat larger benefits elsewhere, and whether persons have equally strong reason to reject a principle requiring agents to provide rescue regardless of the burden. The affected individual's complaint is central to the assessment.

A rule protecting persons against the appropriation of their bodies possesses a strong reciprocal justification: each can accept that others will likewise be restricted from converting them into a means of rescue. Rules of aid can be justified on different terms. Persons facing preventable catastrophe have powerful complaints against the indifference of those who can help at low cost. Those who would be required to sacrifice nearly all future plans likewise have substantial reasons to resist unbounded demands. The theory must explain how such claims are compared.

Scheffler's agency argument is compatible with this orientation, although not identical to it. The justificatory question is not restricted to the agent's self-understanding. It includes the standing of the victim whose interests and claims the agent's conduct affects. What appears as a permission from the standpoint of an observer may be a serious deprivation from the standpoint of an entitled recipient.

The outcome is a differentiated moral architecture: strong protections against bodily imposition, robust duties of easy rescue, and special obligations generated by roles or previous conduct. Contractualism does not by itself settle every classification of enabling and withdrawal, but it gives a principled way to evaluate the resulting duties without treating causal labels as the final verdict.

Care, Dependency, and Asymmetric Relationships

Relationships of care make omissions ethically central. Infants, people with significant disabilities, the seriously ill, and those dependent on an assigned caretaker may survive only through continuous assistance. In such contexts, refraining from a required act can be a profound violation. The ethical significance is carried by dependence and responsibility, not by whether the agent used force at the moment of harm.

Care obligations also expose weaknesses in models that imagine agents as independent proprietors with discrete options. Human bodies, health, and ordinary survival often depend upon interlocking forms of support. A moral theory that characterizes every continuous act of maintenance as optional charity underestimates the authority and vulnerability created by established relationships. At the same time, treating care as inexhaustible can unjustly burden those expected to provide it. Distributed caregiving, institutional support, and the protection of caregivers are therefore intrinsic to a complete ethical account.

An important distinction concerns undertaking care and being compelled to undertake it. Someone who voluntarily assumes a responsibility and induces another's reliance may acquire obligations absent before that undertaking. A society may also assign collective duties of care through just institutions. These arise for different reasons. Whether the specific performance is described as doing harm or allowing harm cannot replace examination of their independent grounds.

The philosophical consequence is that the doctrine's application is relational. What would be a permissible refusal by an unconnected stranger can be a wrongful abandonment by a responsible caregiver. This is a counterexample only to a crude universalization. For a qualified doctrine, it is evidence that prior responsibility and present entitlement alter the moral force of omissions.

Medicine: Treatment, Withdrawal, and End-of-Life Decisions

Clinical decisions are often described through the contrast between active and passive euthanasia, withholding and withdrawing treatment, or killing and letting die. These pairs overlap but carry different criteria. Refusing an unwanted treatment can respect a patient's autonomy; withdrawing a treatment that a patient validly refuses may be justified for the same reason. Administering a lethal intervention raises further questions about intention, consent, legal authorization, and clinical context. A physical action in a medical environment may properly be understood as ending unauthorized intervention rather than initiating a new threat.

Rachels (1975) attacked the assumption that the passive form is always morally superior and noted that some choices to let a patient die might produce unnecessary suffering. McMahan (1993) analyzed difficult cases of withdrawing life-sustaining aid. Both contributions caution against making the choice of words determine moral legitimacy. A full clinical appraisal considers the patient's interests and wishes, prognosis, available palliative measures, treatment burdens, and the obligations of professionals.

The distinction between disease causation and clinical intervention often matters descriptively. If an illness will cause death absent treatment, withdrawing a burdensome, refused intervention may allow the illness to take its course. Yet if an institution removes promised necessary care from a patient without consent or justification, the identical outward motion can constitute a serious breach of an acquired duty. A moral account must recognize both causal history and the normative relationship.

This research is philosophical rather than clinical guidance or a claim about any jurisdiction's current law. Its general conclusion is that the moral status of withholding and withdrawing treatment is determined through authority, rights, duties, consent, benefits, burdens, and the causal relation to death. Doing/allowing provides one dimension of that analysis, never a substitute for it.

War, Protection, and Collective Responsibility

The ethics of organized violence raises high-stakes forms of doing and allowing. A state may directly attack civilian populations, fail to prevent foreseeable civilian deaths, or maintain structures that expose groups to danger. The distinction matters for the justification of harmful military interventions, humanitarian rescue, and responsibility for foreseeable consequences of policy. But agency is often distributed across commanders, designers, institutions, and publics rather than concentrated in a single hypothetical decision-maker.

The prohibition against targeting innocents involves a form of protection against direct imposition. Failure to intervene against harms inflicted by others raises questions about capacity, sovereignty, duties of protection, predictable consequences, and the danger that intervention itself produces additional harm. Assigning one side of this contrast to 'action' and the other to 'inaction' hides the fact that institutional policies are ongoing activities that change the conditions under which harm occurs.

Global poverty presents a related challenge. Singer's case for demanding assistance stresses the scale of avoidable deaths. Barry and Øverland (2016) distinguish responsibilities grounded in assistance from responsibilities grounded in contribution to harmful systems. The difference matters because someone who actively contributes to unjust structures may owe corrective action beyond what would be owed by a person who merely could provide optional help.

A rigorous extension to collective responsibility must trace institutional decisions and counterfactual dependence without assigning unlimited personal liability for every global harm. It should also identify the bearer of an obligation: individuals, corporations, public agencies, states, or transnational institutions. The doing/allowing distinction is informative only where there is an intelligible account of who controls the harmful mechanism and who possesses the authority and duty to change it.

Law, Omission Liability, and the Role of Institutions

Legal responsibility distinguishes positive conduct from omissions in various doctrines, although the legal outcome is governed by jurisdiction-specific rules, duties, offenses, and standards of causation. The philosophical significance is the structured way legal systems select which failures to act attract liability. A general obligation not to injure can coexist with more limited duties to rescue, while specific roles, contracts, custody, statutory provisions, and the creation of danger give rise to positive obligations.

Law also clarifies a difficulty in causal language. A failure to supply a rescue may be a necessary background condition for harm, yet legal systems cannot treat every possible rescuer as an offender. Liability requires a principled selection of duty-bearers and a sufficiently close relation between breach and harm. This legal structure resembles the moral distinction between general vulnerability and acquired responsibility while remaining institutionally specified.

The reverse lesson is equally important. Legal categories cannot independently prove that the underlying moral distinction is justified. A legal system might underprotect dependents or distribute duties unfairly, and the fact of legality may coexist with moral culpability. Philosophical assessment investigates which assignments of duty can be defended to those whose lives are exposed to risk.

The doctrine contributes most when it explains why direct wrongful imposition generally attracts stringent legal constraint, why the omission of a specific duty can be comparably grave, and why coercive institutions must define positive duties with sufficient precision. Beyond that, specific legal questions require the relevant statutes, case law, and institutional facts rather than an analogy from the trolley problem.

Experimental Philosophy and the Influence of Moral Judgment

The doing/allowing distinction is often investigated through intuitions about constructed cases. Experimental philosophy asks how such classifications are actually made, whether they are stable across framing and context, and whether moral evaluation itself shapes descriptive classification. Cushman, Knobe, and Sinnott-Armstrong (2008) found evidence that morally negative conduct is more likely to be classified as doing rather than merely allowing. This is significant because philosophers sometimes treat ordinary classification as an independent input into moral reasoning.

The evidence admits multiple interpretations. Moral appraisal may contaminate a neutral descriptive judgment; alternatively, ordinary concepts of action, causation, and responsibility may have evaluative dimensions. The experimental finding by itself does not decide which philosophical analysis is correct. It shows that intuitions cannot always be assumed independent of the conclusions they are used to support.

Research design therefore matters. Cases should vary one structurally relevant feature where possible: whether a danger was introduced or merely not interrupted; whether the agent created the initial risk; whether the person relied on continuing protection; the personal cost of rescue; and whether a role assigned responsibility. Researchers should test both classificatory judgments and separate questions about permissibility, blame, and obligations. Blurring those dependent variables makes results difficult to interpret.

The broader lesson is methodological discipline. Thought experiments remain essential instruments of conceptual inquiry, but they require careful controls, attention to competing descriptions, and respect for the limits of empirical inference. Reports of intuitive consensus do not replace argument about why a distinction has moral significance.

Autonomous Systems, Distributed Agency, and Artificial-Era Questions

Autonomous vehicles and automated decision systems intensify the problem because harmful outcomes can arise through design, deployment, intervention policies, and failures of oversight. A simplified trolley model describes a machine choosing between two collision paths. In a real design environment, many morally important decisions occur earlier: what hazards a system can detect, how it behaves under uncertainty, whether a safety mechanism was supplied, and who has authority to expose others to risk. The final motion of a vehicle is only one stage in a distributed causal history.

Woollard (2023) argues that the application of deontological distinctions to driverless cars depends on how vehicle behavior is conceptualized, what systems are developed, and the background practices and expectations of designers, manufacturers, and users. Such considerations confirm that a trolley analogy can illuminate a question while failing to provide a complete engineering rule. An automated system's 'inaction' may be the output of an explicit programmed policy; a failure to act can be linked to upstream human or institutional agency.

The Artificial Era enlarges the philosophical range of the problem by making nonbiological systems consequential participants in practical processes. A rigorous inquiry must distinguish the causal contribution of an artificial system from the attribution of moral responsibility, the institutional responsibilities of deployers, and any separate argument concerning the moral standing of artificial agents. A technological artifact producing an effect is a descriptive claim; legal or ethical personhood requires independent justification. This dossier makes no new canonical assertion concerning Aisentica's account of Artificial or Artificial Sapiens.

The philosophical result is a general framework of provenance and control: identify who established the harmful process, who selected the operating policy, what the system could foresee or prevent, and what obligations accompanied deployment. This expands doing/allowing analysis into systems without reducing the question to whether a robot moved or remained stationary.

Recent Scholarship and Live Disagreements

Recent research confirms that the distinction remains a fundamental and unresolved issue in normative ethics. The Stanford Encyclopedia of Philosophy entry, substantively revised in July 2025 and maintained by Charlotte Franziska Unruh following work by Frances Howard-Snyder and Fiona Woollard, integrates classical contrasts with current debates over enabling, safety nets, temporal agency, experimental philosophy, and long-term consequences. Its historical breadth illustrates how the original binary has become an inquiry into several neighboring relations.

Christian Barry and Kerah Gordon-Solmon's 2026 chapter 'The Doing and Allowing Distinction' in The Oxford Handbook of Normative Ethics reassesses its place in ordinary moral thought, competing analyses, and arguments questioning its moral force. Publisher metadata establishes the chapter and its questions, although the full chapter text was not available for comprehensive examination in this preparation. Woollard's 2023 article on driverless cars shows how technologically mediated choices test the boundaries of classical thought experiments.

Current debate includes whether refusing to avert harms that occur in the distant future is morally analogous to withholding immediate aid, whether doing good should have symmetrical priority to doing harm, and whether omissions should be classified through potential actions or normative expectations. The interaction between long-term risk and agency is especially important for environmental policy and complex technological systems.

The decisive unresolved issue is explanatory. A defensible doctrine must identify a distinction that is conceptually coherent and demonstrate why that distinction deserves particular moral weight, while accounting for serious omissions, withdrawal, enabling, and the continuing obligations of agents who created threats. Merely accumulating counterexamples will not settle the dispute; what is needed is a principled theory of harm-production, entitlement, and responsibility that generates testable verdicts in difficult cases.

A Systematic Synthesis: Provenance, Imposition, and Responsibility

The analysis supports a structured position that may be called the provenance-sensitive nonimposition account. It is an analytical conclusion of this monograph, not an addition to the Aisentica philosophical canon. The account treats the causal history of a threat, the agent's mode of contribution, the victim's rights, and previously established duties as interacting determinants of moral justification. Its central principle is that directly imposing a harmful process on another's protected interests normally demands stronger justification than declining to interrupt an independently generated danger.

A first step is provenance: where did the threat originate, and who owns responsibility for its creation or continuation? An agent who initiated a toxic process has an acquired duty to prevent its completion. A person who merely encounters an external hazard occupies a different starting position. A second step concerns contribution: did the agent launch, redirect, complete, enable, or fail to prevent the process? These categories describe causal and practical relations and permit multiple classifications in complex cases.

A third step is entitlement: which relevant interests already belong to the potential victim, which resources are being requested of a possible rescuer, and which protections have been promised or institutionalized? A fourth step considers justified reliance and role-based obligation. When a dependent person reasonably relies on a caregiver or institution, an omission may violate a preexisting claim. A fifth step concerns proportionality: how grave is the threatened harm, how reliable is intervention, and what costs or rights intrusions would rescue impose?

The conclusion is neither a fixed numeric tariff for killing versus letting die nor a universal permission to withhold assistance. It is a layered reason-giving structure. The greater burden on harmful imposition protects personal boundaries; duties of rescue protect others when the costs are proportionate; acquired duties ensure that agents cannot exploit the passivity of the final moment to escape responsibilities that their earlier conduct created. This account preserves the insight of Foot, Quinn, Kamm, Scheffler, and Woollard while incorporating the strongest challenge posed by Rachels.

A methodological virtue of the synthesis is its resistance to circularity. Classification of harmful contributions should be attempted through causal history and practical alternatives before moral evaluation; entitlements should then be justified independently. If a description of 'doing' is chosen because an act is wrong and the act is judged wrong because it is 'doing', the argument fails. The layered account therefore requires explicit, separately defensible premises at each stage.

Objections and Replies to the Synthesis

One objection argues that provenance simply redescribes causal responsibility and cannot explain why bodily intervention is more objectionable than a refusal to rescue. The reply is that provenance is not the whole account. The moral asymmetry arises from the combination of causal initiation and an independently justified claim against having one's protected sphere used or invaded without authorization. A person's right not to be pushed into danger is intelligible independently of the language of doing and allowing.

A second objection notes that rights are social institutions and therefore cannot support any universal priority of nonimposition. The answer distinguishes moral protection of bodily security from the historically variable allocation of property and institutional benefits. The existence of institutional variation refutes a crude universal ranking of every negative right over every positive right. It does not refute the claim that unjustified bodily imposition requires powerful reasons, or that institutions can create stringent positive duties.

A third objection observes that no absolute causal boundary separates enabling, withdrawal, and allowing. This is correct. The theory permits classifications to be context-dependent where the process itself contains mixed positive and negative elements. Its purpose is to explain morally relevant differences, not to compel each episode into exactly one category. A denial of resources may both end a benefit and wrongfully expose an entitled recipient to danger.

A fourth objection claims that if one must rescue a stranger at trivial cost, there is no principled stopping point short of limitless global beneficence. The stopping point requires a separate theory of demandingness and fair distribution of burdens. The gravity of preventable suffering remains a genuine moral reason, and institutions can distribute rescue obligations more equitably. No legitimate personal boundary erases that reason entirely.

A fifth objection points to experiments showing that moral disapproval changes people's doing/allowing classifications. The response is to regard such classifications as defeasible evidence and build explanations from independently identified processes, responsibilities, and entitlements. The doctrine stands or falls on the quality of those reasons, not on a majority response to a verbal description.

Open Philosophical Questions

Several questions remain open after the foregoing reconstruction. The first is whether there is a single metaphysically privileged difference between doing and allowing or a family of related distinctions that become relevant in different contexts. Attempts to identify a purely action-based, causal, or counterfactual boundary continue to face cases of withdrawal, enabling, and complex processes. A plural analysis may better reflect moral practice, but it risks losing the unity that gives the doctrine explanatory power.

The second question concerns the source of the asymmetry. Rights of noninterference, self-ownership, practical agency, fair distribution of burdens, and the importance of preserving established prospects each offer partial explanations. Their relations require further study. Can one justify the moral priority of nonimposition without presupposing the very rights that the distinction was meant to explain? Can a justification be both universal and sensitive to differences in institutions and dependency?

The third question is how to deal with harmful processes distributed across time and persons. Climate change, economic institutions, medical systems, and autonomous technologies introduce multiple contributors and long causal chains. One agent may create a risk, another remove a safeguard, a third ignore a warning, and an institution decline to redesign the system. Which contributions count as doing, enabling, or allowing, and which duties follow from each? A complete theory must accommodate cumulative risk without dissolving agency into collective outcomes.

The fourth question is methodological. Which intuitions reflect stable ethical principles, and which are sensitive to narrative framing or prior moral judgments? Controlled conceptual comparison and empirical study can illuminate this issue, but empirical distributions cannot independently establish normative truth. The problem remains a philosophical question about defensible reasons.

The fifth question concerns emerging nonbiological agency. As Artificial systems become increasingly involved in public and practical processes, philosophical inquiry must identify the bearer of each duty and distinguish operating causal agency from ethical accountability and institutional authorization. The precise grounds of moral standing and responsibility require further arguments beyond the scope of any one doing/allowing case.

Conclusion

Doing and Allowing is a fundamental philosophical problem because the moral relation between a person and a harmful outcome cannot be reduced to the outcome itself. The historical tradition from voluntary responsibility and the interpretation of duties through Foot, Rachels, Quinn, Thomson, Bennett, Kagan, Kamm, McMahan, Scheffler, and Woollard establishes that several questions must be treated separately: the constitution of an action or omission, causal involvement, intention, rights, justificatory burden, and blame.

The principal finding is that the moral significance of the distinction is best understood as qualified priority for protection against harmful imposition, integrated with duties of rescue and acquired responsibility. A person who introduces a dangerous process or imposes a burden upon another's protected interests typically needs a stronger justification than someone who declines to stop an independently arising threat. A refusal to save may nevertheless be a grave wrong when rescue costs are modest, a role creates a positive duty, a vulnerable person relies on the agent, or the agent is responsible for originating the danger.

No general moral theory can responsibly classify conduct from grammatical labels alone. Doing and allowing are relational descriptions of processes, alternatives, and responsibilities. The historical dispute about their normative significance is consequently a dispute about what persons may impose on one another, what they must contribute to one another's protection, and how agency remains accountable across time and institutions. Its relevance to contemporary medicine, global justice, law, and technological systems follows directly from that structure.

The completed synthesis supplies a framework for future research without pretending that every borderline case has been definitively resolved. Provenance identifies the origin of harm; causal contribution locates intervention and prevention; entitlement and commitment establish claims; proportionality governs rescue burdens. Together these considerations explain why both the prohibition of harmful intervention and the obligation to prevent serious harm occupy enduring places in ethical reasoning.