The Principle of Foreseeable Harm to Innocents (FHI): A Doctrinal Framework for the Protection of Ascertainable Civilians in Armed Conflict

FHI JCSL Article Submission.pdf

Download PDF • 703KB

Abstract

Modern targeting doctrine permits the foreseeable killing of ascertainable civilians provided the harm can be characterised as 'unintended' and 'not excessive.' This Article argues that existing international humanitarian law, properly interpreted, already prohibits such harm where feasible alternatives exist. It articulates the Principle of Foreseeable Harm to Innocents (FHI): a doctrinal standard, grounded in Article 57 of Additional Protocol I, that interposes an 'avoidability gate' between the identification of a military objective and proportionality analysis. Attacks may not proceed against foreseeable civilian risk until the attacking authority has demonstrated that no feasible alternative could eliminate or substantially reduce the harm.

The framework builds on Ohlin's defence of the intent/foresight distinction, Crawford's analysis of systemic collateral damage, Shue's capacity-indexed account of legal obligation, and Lewis's operational research on civilian harm detection failures.

The Article addresses tactical and operational targeting decisions - planned strikes where decision-makers have time to assess civilian presence and consider alternatives - not political decisions preceding armed conflict nor dynamic engagements where reaction time forecloses deliberation.

The central contribution is methodological: shifting legal inquiry from reconstruction of subjective intention to verification of objective alternatives. Enforcement becomes a matter of examining what the commander knew and what alternatives existed, not speculating about mental states. This aligns with Article 57's capacity-indexed structure, drafted to expand precautionary obligations as technological capability increases.

FHI does not supplant proportionality; it ensures proportionality addresses only genuinely unavoidable harm. By anchoring legality in verifiable facts rather than contestable mental states, FHI offers a doctrinally grounded mechanism for addressing foreseeable, avoidable harm to ascertainable civilians in deliberative targeting.

I. INTRODUCTION

What legal constraints apply when a planned attack will foreseeably kill civilians who are identifiable in advance, when those civilians are not the object of attack, when proportionality analysis is formally satisfied - and when feasible alternatives exist that would avoid the harm?

This Article addresses that question at the level of tactical and operational targeting decisions - the specific choices made by commanders and targeting cells about whether, when, and how to strike. It does not address the political decisions that precede armed conflict, nor the intentional targeting of civilians, which is unambiguously prohibited. The focus is narrower: planned attacks in which civilians are not targeted but are foreseeably at risk, and in which alternatives exist that would reduce or eliminate that risk.

Contemporary armed conflicts increasingly present a recurrent and troubling pattern: attacks in which civilian harm is not merely possible but foreseen as probable, civilians are not the object of attack, and proportionality analysis is formally satisfied - yet civilians who are identifiable in advance are nonetheless killed. These cases occupy an uneasy doctrinal space. They do not fit the paradigm of intentional targeting of civilians, nor are they readily condemned as disproportionate under prevailing interpretations of international humanitarian law (IHL). And yet the harm they produce is neither accidental nor unavoidable.[1]

From Gaza to Ukraine, from Yemen to Somalia, contemporary conflict repeatedly produces the same tableau: an attack is launched against a military target; civilians in its vicinity are plainly visible; targeting cells assess that harm is 'unlikely but possible' or 'likely but justified'; and the strike proceeds. When the dust clears, the law is invoked not as a restraint but as a shield: the harm was foreseen but 'unintended,' foreseeable but 'not excessive,' tragic but 'lawful’, regrettable but ‘unavoidable.’[2] 

The pattern extends to targeted killing operations. In May 2016, a US drone strike killed Mullah Mansour, then leader of the Taliban, on a road in Pakistan. Also killed was Mohammad Azam, a taxi driver and father of four who had picked up Mansour as a fare six hours earlier at a border crossing. Azam was identifiable, his presence in the vehicle was known, and his death was not accidental but foreseen and accepted. Local officials confirmed he had no connection to any militant group. His family received no compensation; his death was classified as lawful collateral damage. Under current doctrine, it was - despite a civilian casualty rate of fifty per cent.

The central claim advanced here is that existing IHL already contains the resources to answer this problem - but that those resources are systematically under-used. Properly interpreted, the requirements of constant care and feasible precautions under Additional Protocol I, read in light of the Conventions’ humanitarian object and purpose, imply a logically prior inquiry: whether foreseeable civilian harm is avoidable.[3] Where such harm can be avoided through feasible alternatives, proportionality analysis should not operate as a permissive balancing test at all. Instead, avoidability functions as a gatekeeping constraint: if foreseeable harm to identifiable civilians can be prevented, the attack must not proceed.

The stakes extend beyond academic refinement. If the law permits the foreseeable killing of ascertainable innocents under the cover of ‘military necessity,’ it fails its core constituency. Civilians are not abstractions; they are persons with faces, names, and families. When law characterises their deaths as acceptable ‘side-effects,’ it provides not protection but complicity dressed in legal language.

The Principle of Foreseeable Harm to Innocents (FHI) articulated here does not propose a new rule of IHL, nor does it seek to collapse distinction into proportionality.[4] Rather, it clarifies the sequencing of existing obligations. It insists that proportionality analysis is properly triggered only after an attacking authority has discharged its duty to identify and assess feasible alternatives that would avoid foreseeable civilian harm. In this sense, FHI operates not as an expansion of civilian immunity, but as a specification of when proportionality reasoning is legally available at all.

Scope and Limitations

The argument advanced here is deliberately limited. It addresses two categories of operations: (1) planned strikes where civilians might be at risk, and (2) targeted killing operations where civilians might be at risk. In both categories, decision-makers have time and informational capacity to assess civilian presence and consider alternative courses of action. FHI does not purport to regulate deliberative targeting decisions made under rapidly evolving battlefield conditions, where information is fragmentary and reaction time is measured in seconds. ‘Deliberative’ in this context refers to the availability of time and information sufficient to consider alternative courses of action, not to the absence of operational pressure or risk.

Infrastructure and dual-use facilities raise distinct issues that this Article acknowledges but does not fully address. The avoidability principle applies where destruction of civilian objects would foreseeably cause civilian deaths or serious injury - for example, attacks on infrastructure essential to civilian survival. However, the longer-term, indirect effects of infrastructure targeting, and the particular complications of dual-use facilities, deserve sustained analysis that exceeds the scope of the present inquiry.[5]

Some operational analysts emphasise that even well-resourced forces often fail to detect civilian presence in time to prevent harm; FHI accepts that reality and confines its reach accordingly.

Political Choice and Technological Capacity

The analysis that follows treats contemporary intelligence, surveillance, and reconnaissance (ISR) capabilities not as neutral facts, but as capacities whose legal significance depends on how and why they are developed and deployed. ISR - the integrated systems of sensors, platforms, communications, and analysis that enable military forces to observe, identify, and track objects and persons on the battlefield - has transformed what can be known about civilian presence before, during, and after attack. It must be borne in mind, however, that the capabilities themselves are political choices. States choose whether to invest in surveillance systems that make civilian presence visible; they choose whether to develop precision weapons that enable discrimination; they choose how to integrate ISR into targeting processes. FHI holds states accountable for the obligations their own choices create.[6]

The same political dimension pervades targeting decisions. The determination of who counts as a combatant or civilian is shaped by rules of engagement, threat definitions, and political judgments about who may be targeted. The US Standing Rules of Engagement, for example, define 'hostile intent' broadly to include threats to mission accomplishment, and interpret 'imminent' to mean something other than 'immediate or instantaneous.'[7] These definitional choices expand the category of targetable persons and compress the space for civilian protection. FHI cannot reform those choices directly, but it can ensure that when states choose to develop the capacity to see civilians and to spare them, that capacity generates corresponding legal obligation.

Intellectual Foundations

This interpretation has roots in the intellectual tradition from which modern IHL emerged. My own engagement with the law of armed conflict began under the instruction of Professor Colonel Gerald Draper, a central figure in the post-war development of humanitarian law.[8] Draper taught that the legitimacy of IHL depends on its capacity to restrain violence, not merely to provide ex post justifications for it. His teachings on the Nuremberg Trial impressed upon me a principle that runs through FHI: those who possess both the capacity to foresee harm and the means to avoid it cannot disclaim responsibility by invoking the absence of subjective intent. Knowledge, combined with avoidability, generates obligation.

FHI’s intellectual foundations rest on three bodies of work that Part III examines in detail: Ohlin’s defence of the intent/foresight distinction, which shows why foreseeable civilian harm cannot coherently be resolved by reconstructing intent alone;[9] Crawford’s analysis of systemic collateral damage, which exposes the accountability gap intent-focused frameworks produce;[10] and Shue’s insistence that legal obligation must be indexed to operational capacity.[11] FHI synthesises these insights into a single doctrinal mechanism: the ‘avoidability gate.’ Empirical research in moral cognition suggests that this framework tracks deep features of human moral judgment: individuals intuitively distinguish between intended and foreseen harm, and public support for military operations decreases as civilian harm shifts from accidental to foreseeable to intentional.[12]

The structure of the Article reflects the structure of the argument. Part II traces the historical development of civilian immunity and the avoidability principle from medieval origins through Nuremberg to Additional Protocol I. Part III examines the scholarly foundations on which FHI builds. Part IV diagnoses the doctrinal crisis that renders existing law incapable of constraining foreseeable and avoidable harm. Part V constructs the doctrinal architecture of FHI itself, including its relationship to collateral damage estimation methodology. Part VI identifies judicial developments that already point toward avoidability reasoning. Part VII concludes by considering what is at stake when law permits the predictable killing of ascertainable civilians.

II. HISTORICAL DEVELOPMENT OF CIVILIAN IMMUNITY AND THE AVOIDABILITY PRINCIPLE

The core moral commitments underlying FHI - civilian immunity, responsibility for avoidable harm, and the impermissibility of shifting risk onto the innocent - run deeply through the history of jus in bello. What is new is not the principle, but the operational environment: modern ISR technologies now make civilian presence visible, predictable, and avoidable in ways earlier legislators could not imagine.

A. From Medieval Foundations to Nineteenth Century Codification

The moral architecture underlying FHI can be traced from medieval Peace of God movements through the Hagenbach trial (1474) - frequently cited as an early instance of individual accountability under the laws of war - to modern codification.[13] These developments established that certain categories of persons stood outside the legitimate scope of military violence, and that those who harmed them could be held accountable regardless of military necessity claims. The 1863 Lieber Code required that 'the unarmed citizen is to be spared in person, property, and honor as much as the exigencies of war will admit'’[14]; the 1907 Hague Regulations required belligerents to 'do all in [their] power to warn' before bombardment[15]. These codifications transformed moral intuitions into positive legal obligations, establishing the textual foundation for modern precautionary requirements.

B. Nuremberg, Yamashita, and Knowledge-Based Responsibility

The Nuremberg Tribunal (1945-46) consolidated the principle that knowledge of foreseeable civilian harm, combined with the capacity to prevent it, generates legal responsibility regardless of superior orders.[16] The Hostage Case established that a commander who knows subordinates are committing violations and fails to prevent it bears liability - the doctrine of command responsibility.[17] These cases established that foreseeability of harm can substitute for intent in the context of culpability; knowledge of likely consequences is sufficient to trigger legal responsibility.

The Yamashita proceedings reinforced this trajectory.[18] General Yamashita was held accountable for atrocities committed by troops under his general command in the Philippines. The tribunal’s reasoning established that commanders bear an affirmative duty to know what their forces are doing and to prevent foreseeable violations. This duty of knowledge - the obligation to ascertain what can be ascertained - resonates directly with FHI’s insistence that where civilian presence is knowable, failure to know cannot excuse

subsequent harm.

C. Additional Protocol I and Capacity-Indexed Obligations

Additional Protocol I (1977) carried this logic of knowledge-based responsibility into treaty form. Article 48 requires parties to ‘at all times distinguish between the civilian population and combatants.’ Article 51 prohibits attacks ‘expected to cause incidental loss of civilian life… which would be excessive.’ Article 57 requires those who plan attacks to ‘take all feasible precautions’[19] - a capacity-indexed obligation: what the attacker must do depends on what the attacker can do.[20]

The travaux préparatoires and early ICRC commentary repeatedly anchor feasibility in ‘the technical means… available at the time,’ and identify then-emerging reconnaissance and sensing technologies—such as radar, electro-optical imaging, and infrared sensors - as relevant to the duty to identify and verify targets.[21] Subsequent ICRC commentary and state practice have extended this logic to include satellite and other advanced ISR capabilities as they became operationally available. The 1987 ICRC Commentary notes that ‘precautions are prescribed which take into account the developments in visual and other methods of reconnaissance’ and that the duty to verify civilian presence scales with ‘the technical means of identification that are available.’[22]

D. The Fulcrum

The significance of this drafting choice is both profound and determinative - it is the fulcrum of everything that follows. Many legal rules are static: they define a fixed standard that applies regardless of technological change. Article 57 is different. It embeds technological responsiveness into its structure. The feasibility of a precaution is not measured against a 1977 baseline; it is measured against current capability. What was ‘not feasible’ in 1977 may be routine in 2025.

FHI carries this logic forward: where civilian presence is no longer merely foreseeable but positively visible, where avoidability is no longer speculative but operationally real, the precautionary duty thickens into a requirement not to proceed. Far from breaking with AP I, FHI gives interpretive effect to the structural logic reflected in its precautionary framework. The drafters understood what remains true today: IHL must evolve with the battlespace it regulates. A rule of precaution that fails to evolve with technological transformation becomes a rule that effectively permits what it was designed to prevent.

III. SCHOLARLY FOUNDATIONS: INTENT, CAPACITY, AND SYSTEMIC HARM

The Principle of Foreseeable Harm to Innocents does not emerge from a normative vacuum. It rests on three bodies of scholarship that together illuminate a persistent doctrinal failure: the law’s inability to address civilian harm that is foreseen, avoidable, and yet treated as legally incidental. Each addresses a different dimension of the problem, and none alone supplies a complete solution.

A. Intent and the Limits of Moral Reconstruction

The modern law of targeting continues to place significant weight on intention. Civilian deaths are unlawful if civilians are made the object of attack; they may be lawful if civilians are foreseeably harmed but not intended. This distinction, inherited from the Doctrine of Double Effect, has been rigorously defended by Jens David Ohlin, who argues that attempts to collapse foresight into intent - for example through expansive readings of dolus eventualis – fundamentally undermine both moral clarity and legal coherence.[23] If foreseeability alone were sufficient to establish intent, the distinction between distinction and proportionality would dissolve, and the collateral damage rule would lose its conceptual foundation.

Ohlin’s defence is decisive on its own terms. Distinction and proportionality operate as analytically separate tracks, and criminal liability for targeting civilians cannot be reconstructed by smuggling foreseeability into the mental element of intent. To do so would erase the line between lawful and unlawful targeting in ways the law neither supports nor could operationalise. Any doctrinal response to foreseeable civilian harm that depends on collapsing this distinction is therefore bound to fail. This is not a flaw in Ohlin's argument; it is a consequence he explicitly acknowledges.

Yet it is precisely the success of Ohlin’s defence that exposes the problem this Article addresses. If intent and foresight must remain distinct - as Ohlin convincingly argues - then the law cannot rely on mental state to regulate foreseeable civilian harm. Where civilians are not the object of attack, and where proportionality is formally satisfied, the absence of intent effectively forecloses criminal accountability. Foreseeable harm that is not intended is

rendered legally invisible.[24]

B. Systemic Collateral Damage and the Accountability Gap

A different strand of scholarship identifies the consequences of this doctrinal gap at the level of practice. Neta Crawford has shown how contemporary military operations systematically produce civilian harm that is neither accidental nor aberrational, but the predictable result of weapons choices, targeting doctrines, and political decisions.[25] Crawford identifies three categories of collateral damage: ‘genuine accidents; systemic collateral damage, the foreseeable if unintended consequence of rules of engagement, weapons choices, and tactics; and double effect/proportionality killing, accepted as military necessity.’[26]

Her central claim is diagnostic: most civilian deaths arise not from error or excess, but from predictable institutional choices that remain legally insulated. Systemic collateral damage - 'collateral damage [that] has been in a moral blind spot… often seen as an accident, and its occurrence and general incidence as natural and unaffected by policy choices' - constitutes the majority of civilian deaths, yet existing doctrine provides no framework for addressing it.[27]

The significance for targeting law lies in the exposure of an accountability blind spot. When legality turns on intention and proportionality alone, foreseeable harm can be normalised as a cost of doing business. The law permits decision-makers to acknowledge civilian presence, quantify expected deaths, and proceed regardless - so long as the harm is not framed as the object of attack or labelled excessive.

Crawford’s prescription is direct: ‘if we can foresee harm, we would be negligent not to take precautions to avoid the harm even if such harm is not our intention’; moreover, ‘before commanders think about precautions and limiting harm to civilians, commanders must think creatively about avoiding it altogether.’[28] Her contribution lies in exposing the systemic production of civilian harm under intent-centred frameworks. FHI supplies the doctrinal

mechanism her analysis calls for.

C. Capacity-Indexed Obligations and the Ethics of Responsibility

The final foundation concerns the relationship between legal obligation and operational capacity. Henry Shue has long argued that what the law may demand depends on what actors can in fact do. Obligations are indexed to capacity: as the means of prevention and avoidance expand, so too do the responsibilities that accompany them.[29] This insight has particular force in the context of civilian protection, where advances in intelligence, surveillance, and reconnaissance, pattern-of-life analysis, and precision strike capabilities have radically altered what can be known and what can be avoided.

Additional Protocol I reflects this logic. Its precautionary provisions do not impose fixed requirements, but obligations defined by feasibility and constant care - standards that thicken as capacity thickens. What earlier generations of commanders could plausibly claim not to know, or not to avoid, is no longer obscure in many contemporary operations. Civilian presence is often visible, predictable, and persistent over time. Alternatives - delay, capture, cancellation, or different means and methods - are frequently available. A legal framework that treats these facts as morally irrelevant risks freezing obligation at an earlier technological moment.

Max Weber's distinction between the ethics of conviction and responsibility - holding political actors accountable not for pure intentions but for foreseeable consequences - provides additional conceptual support for FHI's core claim, a position that empirical research in moral cognition confirms tracks deep features of human judgment.[30] 

D. Synthesis: Why Avoidability Must Be Prior

Taken together, these three strands point toward a single conclusion. Foreseeable civilian harm cannot be addressed by stretching intent. It cannot be controlled by proportionality alone. And it cannot be treated as inevitable where the capacity to avoid it exists. What is missing is a doctrinal mechanism that asks, before balancing military advantage against civilian harm, whether that harm needed to occur at all.

The Principle of Foreseeable Harm to Innocents supplies that mechanism. It accepts the integrity of the intent/foresight distinction defended by Ohlin, takes seriously Crawford’s demonstration of systemic harm and accountability failure, and gives doctrinal effect to Shue’s capacity-indexed obligations. Its central claim is not that civilian harm must never occur, but that foreseeable harm to ascertainable civilians may be tolerated only where it is genuinely unavoidable. Where feasible alternatives exist, proportionality analysis is not a licence to choose the lesser evil; it is a category error.

E. The Martens Clause as Interpretive Support

The interpretive method underlying FHI finds additional support in the Martens Clause. That provision, present in the Hague Conventions and reaffirmed in Additional Protocol I, establishes that in cases not covered by specific treaty rules, civilians and combatants remain protected by ‘the principles of humanity and the dictates of public conscience.’[31] The Clause operates as an interpretive default: where treaty text is ambiguous or silent, the law favours protection over permission. It forecloses the argument that what is not explicitly prohibited is thereby authorised.

IV. THE DOCTRINAL CRISIS: FORESEEABILITY WITHOUT CONSTRAINT

International humanitarian law’s civilian-protection framework developed in an operational environment in which commanders lacked real-time information about civilian presence and had limited alternatives for minimising harm. In that world, subjective assessments of intention and broad, attacker-defined conceptions of ‘military advantage’ made functional sense. Many of these assumptions are no longer reliable.

A. Capacity, Choice, and the Myth of Technological Neutrality

Contemporary targeting practice is frequently described as a function of technological capability - ISR makes civilian presence visible, pattern-of-life analysis renders it predictable, precision-guided munitions narrow blast effects, and CDE methodologies quantify expected harm. These developments are often treated as neutral operational facts.

In some targeting doctrines - particularly those governing strategic attack - the harm inflicted on civilian populations and infrastructure is not merely tolerated but implicitly instrumentalised, framed as a means of coercion or degradation rather than as an incidental side-effect. FHI rejects this doctrinal slippage categorically: where civilian harm becomes a functional objective of attack planning, the distinction principle has already failed.

That framing is incomplete. The capacities that render civilian harm foreseeable and avoidable are the product of deliberate political and institutional choice. States decide whether to invest in persistent ISR, whether to integrate pattern-of-life analysis into targeting cycles, whether to develop or procure precision weapons that enable discrimination between means and methods, and how much risk to accept in the conduct of operations. They also decide how those capacities are operationalised.

International humanitarian law has never treated capacity as morally irrelevant. The precautionary obligations codified in Additional Protocol I are explicitly indexed to feasibility. What can reasonably be demanded of an attacking force depends on what it can in fact do. As operational capacity expands, so too does the legal obligation. To treat advanced ISR and precision-strike capabilities as legally inert is therefore not neutrality but abdication. It allows states to rely on enhanced capabilities to justify attack accuracy while disavowing the correlative obligation to avoid foreseeable harm that those same capabilities make preventable.

This asymmetry is particularly pronounced in planned operations. Where surveillance has been ongoing for days or weeks, where civilian routines are known, and where targeting cells revisit strike options repeatedly, it becomes implausible to characterise resulting civilian deaths as accidental or unavoidable. The doctrinal crisis extends beyond sequencing failures to the very concept of intent. Some targeting doctrines explicitly contemplate effects on civilian populations as strategic objectives. US Air Force Strategic Attack doctrine identifies ‘switching from attacks on purely military targets to attacks on dual-use infrastructure (civilian infrastructure supporting military functions)’ as a mechanism of ‘escalation dominance,’ and describes ‘creating unrest among an enemy’s population’ as a legitimate coercive effect.[32] 

When effects on civilian morale and population psychology are explicitly intended outcomes of targeting decisions, the claim that civilian harm is merely ‘foreseen’ rather than ‘intended’ loses much of its force.

B. The Indeterminacy of Proportionality

Once intent is correctly confined to the deliberate targeting of civilians, proportionality becomes the principal doctrinal mechanism through which foreseeable civilian harm is regulated. Article 51(5)(b) of AP I prohibits attacks ‘which may be expected to cause incidental loss of civilian life, injury to civilians, damage to civilian objects, or a combination thereof, which would be excessive in relation to the concrete and direct military advantage anticipated.’[33] Where civilians are not the object of attack, legality turns not on mental state but on whether the anticipated incidental harm is excessive in relation to the concrete and direct military advantage anticipated. In contemporary practice, proportionality therefore functions as the law’s residual regulator of civilian harm.

This role, however, exposes proportionality to demands it was never designed to meet. Proportionality is a balancing test, not a necessity constraint. It asks only whether harm is too much, not whether it needed to occur at all. As long as some military advantage can be articulated, and as long as civilian harm does not cross an indeterminate excessiveness threshold, proportionality analysis permits attacks to proceed even where harm is foreseen as probable.

The subjectivity and indeterminacy of the proportionality standard are well rehearsed.[34] Military advantage may be defined narrowly (neutralising a particular target) or broadly (deterrence, force protection, operational momentum). The 2023 DoD Law of War Manual includes 'the security of the attacking force' and 'progressive and systematic achievement of military goals'[35] - a definition under which virtually any strike contributes to military advantage. Civilian harm, by contrast, is treated probabilistically and bounded through modelling assumptions rather than confronted as a determinative constraint. The resulting calculus admits outcomes within which most decisions can be justified ex post.

This creates a structural asymmetry at the heart of proportionality analysis: contemporary ISR and collateral damage estimation methodologies enable attacking forces to quantify expected civilian harm with considerable precision, while ‘military advantage anticipated’ remains as subjective, elastic, and difficult to verify as it has ever been. One side of the proportionality equation can now be measured; the other cannot.

What proportionality cannot do - and does not purport to do - is require the selection of a less harmful alternative where one exists. An attack that foreseeably kills civilians may be proportionate even if a different course of action would have avoided those deaths altogether. Delay, cancellation, capture, or a change in means or methods may be acknowledged as possibilities, but proportionality provides no doctrinal basis for insisting that they be taken. The lawfulness of the attack turns on balance, not on avoidance.

The asymmetry Crawford identifies - ever more precise quantification of civilian harm alongside persistent indeterminacy of ‘military advantage’ - is not a technical failure but a structural feature of proportionality reasoning as currently operationalised.

C. The Standing Rules of Engagement Problem

US Standing Rules of Engagement define 'hostile intent' to include threats to mission accomplishment and interpret 'imminent' as something other than immediate.’^[36] In some areas of operation, US forces have applied targeting criteria that treat all local military-age males as presumptively hostile - effectively inverting Article 50(1)’s requirement that doubt be resolved in favour of civilian status.^[37]  These definitional choices contract the space for civilian protection. The significance for FHI is direct: the emergency override articulated in Part V cannot be grounded in SROE definitions of imminence, self-defence, or mission preclusion. To permit such grounding would collapse the distinction between exceptional catastrophe prevention and routine operational judgment.

D. Intent, Foresight, and the Accountability Gap

Part III.A established why the intent/foresight distinction must be preserved: attempts to collapse foresight into intent destabilise the analytical architecture on which targeting law depends.

The accountability gap emerges at this precise junction. Civilian deaths that were known in advance to be likely, and that could have been prevented through feasible alternatives, are retrospectively characterised as tragic but lawful. Proportionality supplies the justificatory vocabulary; the absence of intent supplies immunity from criminal sanction. The law condemns massacre and excuses accident, but struggles to address the predictable killing of ascertainable civilians that lies between those poles.

FHI is not a rejection of Ohlin’s framework, but a response to the space it leaves deliberately unfilled. It accepts the two-track architecture of distinction and proportionality, and it preserves the intent/foresight distinction intact. What it supplies is a logically prior inquiry that Ohlin’s analysis does not purport to answer: whether such harm to civilians was avoidable in the first place. By locating constraint at the level of precaution and necessity rather than mental state, FHI addresses foreseeable civilian harm without destabilising the doctrinal structure Ohlin has so powerfully defended.

The consequences for international criminal law are direct. As Ohlin has pointed out, the ICTY's reasoning in Galić - treating knowledge of probable civilian harm as sufficient for conviction - effectively collapsed the distinction between targeting civilians and foreseeing their deaths.[38] Subsequent jurisprudence has struggled to operationalise this standard without either acquitting commanders who knowingly caused civilian deaths or convicting them on a theory that erases the line between distinction and proportionality.

The result is a prosecutorial and judicial impasse: intent-based frameworks cannot reach foreseeable harm, while knowledge-based expansions destabilise the doctrinal architecture. This threatens the doctrinal stability of targeting law as a whole. FHI offers a way out of this impasse by relocating constraint from mental state to objective alternatives.

E. Risk Transfer and the Displacement of Harm

These doctrinal features facilitate a further distortion: the systematic transfer of risk away from attacking forces and onto civilian populations. Michael Walzer’s contribution to just war theory is central here. In Just and Unjust Wars, Walzer argues that combatants bear a moral duty to accept risk rather than transfer it to civilians: ‘if saving civilian lives means risking soldiers’ lives, the risk must be accepted.’[39]

This pattern has been described as “risk transfer,” in which the burdens of violence are systematically shifted from combatants to civilians rather than mitigated through alternative courses of action.[40] 

International humanitarian law has long rejected the idea that civilian harm is an acceptable substitute for military inconvenience. Yet without an explicit avoidability inquiry, the law lacks a doctrinal mechanism to enforce that principle in cases where harm is foreseeable but not excessive. Risk transfer occurs not because the law endorses it, but because it fails to prevent it.

F. The Missing Inquiry

Taken together, these features reveal a structural omission in contemporary targeting law. The law asks whether harm was intended and whether it was excessive. It does not ask, in any systematic way, whether foreseeable harm to identifiable civilians could have been avoided.

That omission was less consequential and more sustainable when civilian presence was often uncertain and alternatives limited. It is no longer sustainable in operational contexts where surveillance is persistent, civilian routines are known, and targeting decisions are revisited over extended planning cycles. In such circumstances, the failure to inquire into avoidability permits the predictable killing of civilians to be normalised as lawful collateral damage.

According to one detailed operational analysis, most civilian casualties arise not where collateral concerns are detected and strikes proceed regardless, but where concerns are not detected at all - where harm was objectively foreseeable given available ISR capabilities, but subjectively unforeseen because those capabilities were not deployed, consulted, or integrated into the targeting cycle.[41] 

FHI addresses this gap indirectly through its capacity-indexed logic. Where a military possesses ISR capability that would render civilian presence visible, the systematic choice not to employ that capability cannot excuse harm that would otherwise have been foreseeable. The duty of 'constant care' under Article 57 implies an obligation to look, not merely an obligation not to harm what happens to be seen.[42] A targeting process that fails to task available surveillance assets, or that silos intelligence from collateral damage assessment, cannot claim that resulting civilian harm was unforeseeable simply because no one looked.

This does not mean FHI captures all civilian harm. A substantial category of casualties arises from genuine battlefield uncertainty, from detection failures that no reasonable process would have prevented, and from dynamic engagements where the avoidability inquiry cannot meaningfully apply. FHI addresses the category of harm where presence is ascertainable and harm foreseeable; it does not purport to regulate the broader field of casualties arising from uncertainty, error, or operational constraints that genuinely preclude deliberation. A comprehensive approach to civilian protection requires action on multiple fronts - including investment in civilian-protective ISR, integration of harm-mitigation into doctrine and training, and reform of the structural factors that produce systemic collateral damage.[43] FHI supplies one doctrinal tool within that broader effort, not a complete

solution.

If international humanitarian law permits the foreseeable killing of ascertainable civilians whenever that harm can be characterised as unintended and not excessive, it risks becoming, in Gerald Draper’s words, “a mere cover for the violence it was designed to restrain”.

The doctrinal resources to address this omission already exist. The precautionary obligations of Additional Protocol I require ‘constant care’ and the taking of ‘all feasible precautions’ to spare civilians. What is lacking is not law, but sequencing: a recognition that where harm is foreseeable and avoidable, proportionality analysis should not begin at all.

V. THE PRINCIPLE OF FORESEEABLE HARM TO INNOCENTS

The preceding Parts have established that the traditional structure of international humanitarian law no longer adequately reflects the operational realities of modern armed conflict. This Part now undertakes the constructive task: it articulates the doctrinal architecture of FHI, provides operational definitions capable of judicial application, establishes a targeting rule consistent with both humanitarian protection and military feasibility, and sets out a narrow emergency exception confined to conventional armed conflict.

A. Definitions and Threshold Conditions

The doctrinal construction of FHI begins with precise definitions. These are not innovations ex nihilo but elaborations that are consistent with the logic of AP I, the Martens Clause, and jurisprudence from the ICTY, ICC, and ECtHR.

Ascertainable civilians: Persons who are or can be positively identified as civilians using ISR, observation, documentation, contextual indicators, or other available means, including persons presumed civilian under Article 50(1) of AP I.

Foreseeable harm: Harm that is reasonably predictable given the attacker’s actual intelligence, surveillance, reconnaissance, and analytical capabilities. Harm is foreseeable when a competent targeting cell, exercising good-faith analysis, would assess it as a probable consequence of a proposed attack.

Feasible alternative: A course of action that would avoid or substantially reduce foreseeable civilian harm, is operationally available, and does not require clearly disproportionate or suicidal risk. Feasibility is assessed in light of operational circumstances, including resource constraints and competing demands.[44] Feasibility excludes alternatives requiring clearly disproportionate risk or operational collapse.

Avoidable harm: Foreseeable harm that could be prevented through a feasible alternative.

Unavoidable harm: Foreseeable harm that persists despite good-faith consideration and exhaustion of feasible alternatives.

The triggering condition for FHI is twofold:

(1) civilians must be ascertainable, and

(2) harm to them must be foreseeably probable.

Where these conditions are met, FHI requires the attacking authority to determine, before proceeding to proportionality analysis, whether the foreseeable harm is avoidable.

This triggering condition confines FHI to a defined operational domain. The avoidability gate does not purport to regulate split-second battlefield engagements, where information is fragmentary and alternatives illusory. It applies where decision-making is deliberative: where hours, days, or weeks of planning permit reassessment of timing, means, and method.

B. The Avoidability Gate: A Structured Two-Stage Test

Once triggered, the avoidability gate requires a good-faith assessment through a structured two-stage test.

Stage 1 - The Elimination Test. Can the foreseeable harm to ascertainable civilians be eliminated entirely through a feasible alternative? This stage requires good-faith consideration of a range of options such as:

• Delay: Could the strike be postponed until civilians are no longer present? This includes temporal scaling choices: waiting hours, days, or longer where the military objective is not time-critical, even if such delay entails loss of tactical opportunity.

• Cancellation: Could the military objective be achieved without this strike?

• Capture or ground operation: Could the objective be achieved through non-lethal means?

• Continued surveillance without strike: Could patience produce a lower-risk opportunity?

If the answer to any of these questions is yes, and the alternative is feasible in the sense defined above, the attack may not proceed. Harm that can be eliminated is impermissible.

Stage 2 - The Substantial Reduction Test. If Stage 1 is satisfied - that is, harm cannot be eliminated entirely - the inquiry proceeds to Stage 2: Can foreseeable harm be substantially reduced through a feasible alternative? This stage requires good-faith consideration of:

• Alternative weapons or delivery systems with smaller blast effects

• Different timing, angle, or method of attack

• Different force packages or strike configurations

• Warnings, evacuation, or protective measures for civilians

If a harm-reducing alternative is feasible, the attack may not proceed unless that alternative is adopted. Where multiple alternatives exist, the least harmful feasible option must be selected.

The avoidability gate does not add a new precautionary duty; it determines when precautionary failure renders proportionality legally unavailable.

C. FHI and Collateral Damage Estimation Methodology

The distinction between FHI and prevailing collateral damage estimation (CDE) methodologies warrants emphasis. CDE frameworks quantify anticipated civilian harm against thresholds calibrated to perceived military value, assigning numerical values to expected casualties and establishing approval authorities indexed to those values.[45] These frameworks are designed to manage harm, not interrogate its necessity, normalising civilian casualties as expected costs so long as they fall within authorised parameters.[46] 

Civilian casualty cutoff values - the thresholds at which higher-level approval is required or strikes are prohibited - are themselves a revealing indicator of this approach. These values fluctuate across operations and commands, and their variation underscores that CDE methodology treats civilian harm as a variable to be managed rather than a constraint to be observed.[47]

FHI does not deny the utility of CDE methodologies as internal control mechanisms. It insists, however, on a logically prior question: whether anticipated harm must occur at all. Where feasible alternatives exist - including delay, altered timing, different means, or abandonment - proportionality analysis is legally premature. CDE asks how much harm is acceptable; FHI asks whether harm is avoidable. FHI does not displace CDE; it disciplines it by requiring that avoidability be addressed before any balancing can lawfully occur.

Crawford's critique makes clear this is a systems-design problem: casualty-mitigation algorithms are engineered to manage harm, not prevent it. FHI implies a reordering of

decision logic such that avoidability inquiry precedes harm-quantification.

A clarification regarding Civilian Harm Mitigation (CHM) doctrine is warranted. The DoD distinguishes sharply between IHL - binding legal constraints - and CHM, which represents policy that may exceed legal minima. FHI operates at the level of legal obligation, not operational policy. It argues that properly interpreted IHL obligations already mandate much of what CHM frameworks treat as discretionary.[48]

This framing has practical significance. If avoidability is merely policy, states can invoke operational constraints to contract its application.[49] If it is law - grounded in Article 57's requirement that those who plan attacks 'take all feasible precautions' - then its application is reviewable by the courts. FHI's contribution lies in demonstrating that the legal foundation already exists.

TABLE 1: THE AVOIDABILITY GATE (SUMMARY)

Stage	Question	Legal Consequence
Threshold	Are civilians present or reasonably ascertainable?	If no → standard targeting rules apply. If yes → proceed to Stage 1.
Stage 1: Elimination	Can foreseeable civilian harm be eliminated through delay, cancellation, capture, or continued surveillance?	If yes → attack may not proceed. If no → proceed to Stage 2.
Stage 2: Substantial Reduction	Can foreseeable harm be substantially reduced through alternative weapons, timing, angle, method, warnings, or evacuation?	If yes → attack may not proceed unless harm-reducing alternative adopted. If no → harm genuinely unavoidable; proportionality applies.

 The relationship between FHI and existing targeting methodology is illustrated in Figures 1 and 2.

Figure 1 depicts the avoidability gate within the targeting process: where civilian presence is not foreseeable, the attack may proceed through standard channels; where it is foreseeable, the avoidability gate is triggered. Only if no feasible alternatives exist does the attack proceed to proportionality analysis. Where alternatives are available, the legally required outcome is to delay, modify, or cancel the attack.

Figure 2 compares the standard CDE methodology with the FHI approach, showing how the avoidability gate interposes between the identification of collateral concerns and proportionality assessment. The critical difference is that under FHI, proportionality analysis is available only where harm is genuinely unavoidable.

The avoidability inquiry is not one factor among many. It is a gate that precedes all others. Where foreseeable harm to ascertainable civilians can be avoided through feasible alternatives, the attack is impermissible regardless of how favourable a subsequent proportionality calculation might appear.

This sequencing matters. If avoidability were merely another consideration within proportionality analysis, it would be vulnerable to the same indeterminacy that afflicts proportionality itself. Alternatives could be acknowledged and then discounted by reference to expansive conceptions of military advantage. By contrast, treating avoidability as a threshold constraint ensures that proportionality does not operate as a licence to select a harmful option where a less harmful one was available.

Only where the avoidability gate is passed - only where harm is genuinely unavoidable despite good-faith consideration of alternatives - does proportionality analysis become legally available. At that stage, incidental harm may be weighed against military advantage in the familiar manner. Before that stage, proportionality has no work to do.

E. Avoidability, Precaution, and the Sequencing of Proportionality

A common objection is that avoidability-based constraints merely restate what AP I already requires. That objection misapprehends the function of the inquiry. FHI specifies the logical conditions under which proportionality reasoning becomes legally available at all. Article 57 requires feasible precautions to avoid harm; Article 51(5)(b) governs permissibility once an attack proceeds. If proportionality may justify an attack where Article 57 would have required avoidance through feasible alternatives, the precautionary regime collapses into procedural formality.

This sequencing matters because proportionality analysis is structurally incapable of requiring avoidance. Proportionality weighs anticipated civilian harm against expected military advantage; it cannot ask whether the harm must occur at all. Treating avoidability as a threshold inquiry ensures that proportionality does not operate as a permissive licence to select a harmful course of action when a less harmful one was available. Only once harm is genuinely unavoidable does proportionality analysis have doctrinal work to do.

Framed in this way, the avoidability gate does not expand civilian immunity or create new substantive prohibitions. It gives effect to obligations the law already imposes but rarely enforces in sequence. The result is not a stricter law, but a more intelligible one - one in which precaution and proportionality perform distinct, non-redundant functions within a coherent decision-making structure.

The capacity-indexed interpretation of Article 57 advanced here will face predictable resistance.[50] Military legal advisors in some states have historically resisted any interpretations that expand precautionary obligations in line with technological capability, preferring fixed standards. The objection takes several forms: that capability-indexed obligations impose moving-target requirements that undermine planning certainty; that they disadvantage technologically advanced states relative to adversaries with fewer scruples; that they conflate what forces can do with what law requires them to do; and that they invite judicial second-guessing of operational decisions by reference to capabilities that hindsight makes visible.

These objections deserve response. First, as we have seen, Article 57's drafters deliberately chose a capability-indexed structure. 'All feasible precautions' was never intended as a frozen baseline; the travaux préparatoires and ICRC commentary repeatedly anchor feasibility in 'the technical means available at the time.' To resist capability-indexing is to resist the treaty's explicit textual choice.[51] Second, the concern about disadvantaging advanced militaries inverts the moral logic of civilian protection. The capacity to discriminate is precisely what distinguishes lawful warfare from indiscriminate violence; a state that develops the means to spare civilians cannot coherently claim exemption from the obligation to use those means.[52] Third, the distinction between capability and obligation is precisely what FHI maintains. The framework does not demand that states acquire capabilities they lack; it demands that they use the capabilities they have chosen to develop. Fourth, avoidability review is less susceptible to hindsight distortion than proportionality, precisely because it asks factual questions - what alternatives existed, what was documented - rather than requiring courts to reconstruct the 'correct' valuation of military advantage.[53]

None of this makes adoption certain or easy. States that prefer permissive targeting frameworks will resist interpretations that constrain operational flexibility. The question is whether legal doctrine should accommodate that preference or insist on the constraints the law already implies. FHI argues for the latter - not because it guarantees compliance, but because it provides a standard against which conduct can be measured and accountability pursued.

F. Why Avoidability Is More Justiciable Than Proportionality

A persistent objection to enhanced civilian protection standards is that they invite judicial second-guessing of military judgment. The objection has force against proportionality, where courts must weigh incommensurable values - civilian lives against military advantage - without the expertise or situational awareness commanders possess. It has far less force against avoidability.

The avoidability inquiry does not require courts to value military advantage. It asks a different question: did feasible alternatives exist? This is not a scalar judgment about whether harm was 'excessive' but a binary inquiry about whether options were available. Was delay possible? Could a different weapon have been used? Was capture considered? These questions admit of evidential answers. They do not require judges to second-guess tactical valuation; they require judges to verify whether alternatives were explored and why they were rejected.

Courts are well-practised in this mode of review. Human rights jurisprudence routinely asks whether less harmful means were available; administrative law requires decision-makers to consider alternatives and provide reasons for rejecting them. Unlike proportionality, which requires assigning relative weight to incommensurable goods, avoidability asks factual and procedural questions courts regularly handle. The inquiry proceeds independently of how standing rules of engagement define 'hostile intent' or 'imminence.'

G. Proportionality After Avoidability

Once the avoidability gate has been passed, proportionality analysis proceeds in its familiar form. FHI does not assert that civilian life always trumps military objectives. It asserts only that foreseeable civilian harm may not be accepted where that harm could have been avoided altogether.

Nothing in the Principle of Foreseeable Harm to Innocents alters the content of proportionality as such. FHI does not redefine excessiveness, does not privilege civilian life over military advantage in all cases, and does not require judicial hindsight to supplant operational judgment. It preserves proportionality’s role as a balancing mechanism where balancing is unavoidable.

What FHI rejects is the inversion of that role. Proportionality is not the primary regulator of foreseeable civilian harm, but a residual one. It cannot perform its function coherently unless it is confined to cases in which harm was genuinely unavoidable. When proportionality is asked to justify harm that could have been prevented altogether, it is forced to do work it was never designed to do, and predictably fails to constrain.[54] 

What FHI addresses is the prior question of sequencing: ensuring that proportionality operates as a residual mechanism for genuinely unavoidable harm, not as a permissive licence for harm that could have been prevented. The complex interplay of legal doctrine, operational practice, and technological capability cannot be resolved by any single doctrinal intervention. FHI supplies one element of a necessarily multi-faceted response.

H. Risk Allocation and the Burden of Protection

The avoidability gate has implications beyond doctrinal sequencing: it reconfigures how risk is allocated between combatants and civilians. Under prevailing practice, once an attack is framed as directed at a lawful military objective and cleared through proportionality analysis, the residual risk of civilian harm is treated as an unfortunate but acceptable consequence of military necessity. Where the law imposes no obligation to avoid foreseeable harm, civilians become the default bearers of residual risk.

FHI does not deny that military operations involve risk, nor that commanders must weigh competing dangers under uncertainty. What it insists upon is that where civilian harm is foreseeable and avoidable, the law requires that risk be borne by the attacking force rather than displaced onto civilians. Delay, increased exposure, loss of opportunity, or additional operational complexity are not pathologies of lawful warfare; they are the costs of compliance with civilian protection obligations.

This insistence is not radical. It follows directly from the precautionary logic of Additional Protocol I. The duty of constant care is meaningless if it does not sometimes require foregoing the most expedient option. The obligation to take all feasible precautions cannot be satisfied if feasibility is defined so narrowly that it excludes any alternative involving increased risk to combatants. FHI treats capacity as obligation-generating: a force that chooses to develop the means to see civilians and to spare them cannot lawfully ignore that capacity when planning attacks.

I. The Emergency Override

The Principle of Foreseeable Harm to Innocents recognises a narrow emergency override. In exceptional circumstances, foreseeable harm to civilians may be permissible where it is the sole means of preventing vastly greater, imminent harm to civilians. The override exists to preserve the law’s moral credibility in catastrophic edge cases, not to license routine trade-offs.

The conditions are cumulative and exacting: (1) a specific mass-casualty attack against civilians is imminent - deaths in the tens, hundreds, or thousands, within a timeframe that forecloses alternatives; (2) no feasible alternative exists; (3) foreseeable harm is genuinely unavoidable as determined through the two-stage test; (4) expected casualties from the strike are substantially lower than those from the imminent threat - a civilian-to-civilian comparison, not a military-advantage calculus; (5) intelligence is current, corroborated, and highly reliable; (6) the attack is authorised at the highest level practicable; and (7) the commander documents the factual basis contemporaneously, triggering automatic post-strike review.

These conditions sharply distinguish the emergency override from doctrines of self-defence, necessity, or proportionality. They are triggered only where the failure to act would itself constitute a violation of civilian-protection obligations.

J. Scope Limitations: Conventional Armed Conflict Only

The override is strictly confined to conventional armed conflict. It does not extend to nuclear weapons - the logic of catastrophic civilian rescue cannot extend to weapons whose use would itself constitute the catastrophe the override exists to prevent. Nor does it extend to strategic deterrence or anticipatory uses of force. The override cannot be grounded in SROE concepts of imminence or self-defence; importing those interpretations would collapse the distinction between exceptional catastrophe prevention and routine operational judgment.

This exclusion is not incidental. The emergency override is inspired by Michael Walzer’s concept of ‘supreme emergency,’ which Walzer himself linked to nuclear use.[55] Henry Shue has provided a powerful critique of supreme emergency reasoning, demonstrating its vulnerability to abuse and its tendency to license precisely the catastrophic harm it purports to prevent.[56] By confining the override to conventional tactical operations and excluding nuclear weapons entirely, FHI prevents catastrophic-rescue logic from being captured by strategic arguments that would justify mass civilian harm in the name of speculative future prevention.

The override is a residual moral doctrine of catastrophic civilian rescue. Its rarity is a feature: frequent invocation signals misapplication.

K. Human Shields and Responsibility Allocation

Any doctrinal framework centred on foreseeability and avoidability must confront the problem of human shields. Without careful treatment, an avoidability-based approach risks creating perverse incentives: defenders could render military objectives immune from attack simply by surrounding them with civilians, and attackers could invoke that presence to absolve themselves of precautionary obligations.[57]

International humanitarian law unequivocally prohibits the use of civilians to shield military objectives.[58] The deliberate placement or retention of civilians in proximity to military targets in order to deter attack constitutes a serious violation of the law of armed conflict.[59] Where such conduct occurs, primary responsibility for the creation of civilian risk lies with the shielding party.

FHI fully incorporates this principle. The presence of human shields does not convert civilians into lawful targets, nor does it extinguish the attacker’s obligations of distinction, precaution, and proportionality. But neither does it permit the shielding party to exploit civilian presence to immunise military objectives from lawful attack.

The existence of human shielding does not negate the avoidability inquiry. Where civilians are foreseeably present - whether voluntarily or involuntarily - and where harm to them could be avoided through feasible alternatives, the avoidability gate remains operative. The attacker cannot simply point to the defender's unlawful conduct and proceed as though all precautionary obligations have been discharged. The presence of human shields reallocates responsibility but is relevant to the feasibility assessment, not conclusive of it. Where shielding makes delay, capture, or alternative means genuinely impracticable, the avoidability gate may be passed and proportionality analysis may proceed. Where alternatives remain feasible despite the shielding - postponement until civilians disperse, continued surveillance, or non-lethal disruption - FHI requires that those alternatives be taken.

Operational experience suggests that mitigation options in human-shielding scenarios are more numerous than targeting doctrine typically acknowledges.[60] These may include extended surveillance, precision engagement, calibrated warnings, non-kinetic effects, or ground operations. The avoidability gate requires good-faith exploration of these options before concluding that harm to shielded civilians is genuinely unavoidable.

The distinction between 'voluntary' and 'involuntary' shields carries legal significance for responsibility allocation, but FHI's avoidability inquiry proceeds regardless of how civilian presence is characterised: the question is whether harm can be avoided, not whether civilians can be blamed for remaining.[61] 

The human shields problem illustrates the inherent difficulty of situations where both parties may be violating their obligations. The defender who shields violates Article 51(7); the attacker who strikes without exhausting alternatives may violate the precautionary requirements of Article 57. FHI’s treatment of this problem is 'as good as it can be', given the inherent conundrum that both sides could be violating their obligations.[62]

L. Enforcement Architecture

No legal doctrine is viable unless it can be enforced. A persistent difficulty with proportionality has been its dependence on a commander’s internal mental state, rendering post hoc inquiry almost impossible. FHI addresses this difficulty by relocating legal assessment from intention to evidence. FHI does not impose liability for civilian death as such, but for the failure to avoid foreseeable harm where avoidance was feasible

Burden-shifting. When civilians are foreseeably harmed, the attacker bears an obligation to account for the strike’s legality. Article 57’s requirement of ‘all feasible precautions’ and ‘constant care’ implicitly demands that attackers be able to demonstrate compliance - a duty that cannot be discharged without accounting for what alternatives existed and why they were rejected.[63]

Documentation. Modern militaries already necessarily generate extensive records: ISR feeds, sensor logs, targeting packets, command authorisations, collateral damage estimates, and post-strike assessments. FHI requires the preservation of these materials and prohibits tampering, transforming them into a legal substrate for post-strike review.

Reviewability. Courts, commissions of inquiry, and fact-finding bodies can evaluate foreseeability, avoidability, and alternatives through objective data. Was the civilian presence visible on ISR? Were alternatives considered and documented? Was the strike authorised despite the availability of harm-reducing options? These questions are evidentiary, not psychological. They can be answered by reference to records, not speculation about mental states.

VI. JUDICIAL SIGNALS: EMERGING AVOIDABILITY LOGIC

The Principle of Foreseeable Harm to Innocents does not require courts to invent a new category of civilian protection. Across multiple judicial contexts, courts have already displayed sensitivity to questions of foreseeability, precaution, and alternative courses of action, even where the doctrinal vocabulary of ‘avoidability’ is absent. What is emerging is not a new rule, but a pattern of reasoning that increasingly treats the preventability of harm as legally salient.

A. Human Rights Jurisprudence and Operational Planning

The jurisprudence of the European Court of Human Rights provides the clearest illustration of this shift. In cases arising under the right to life, the Court has consistently emphasised that the legality of lethal force depends not only on the moment of execution, but on the planning and control of the operation as a whole.

In McCann v United Kingdom, the Court required states to take feasible steps to minimise foreseeable harm when deploying lethal force.[64] In Finogenov v Russia and Tagayeva v Russia, the Court examined whether viable alternative tactics or methods were considered.[65] In Isayeva heavy free-falling bombs in a populated area were held unjustified when alternatives were available.[66]

Although framed in human rights rather than IHL terms, this jurisprudence reflects a logic closely aligned with avoidability. The Court’s focus is not confined to whether force was excessive at the moment it was used, but whether the state organised its operations so as to reduce foreseeable risks to life. Failures to adjust timing, to delay engagement, or to explore less harmful alternatives have been treated as legally relevant omissions rather than unfortunate by-products of lawful action.

B. International Criminal Law and Command Responsibility

International criminal tribunals operate under stricter constraints. Criminal liability remains anchored in intent, and foreseeability alone cannot substitute for the mental element required for conviction. Nonetheless, judicial reasoning in this context has repeatedly acknowledged that foreseeability bears on assessments of reasonableness, precaution, and command responsibility.

In Prosecutor v Galić, the ICTY treated knowledge of likely civilian harm as determinative in assessing the lawfulness of attacks on Sarajevo.[67] In Prosecutor v Kupreškić, the tribunal examined whether commanders took all reasonable measures to prevent or minimise harm.[68] The ICC in Prosecutor v Ntaganda examined whether alternatives were available and whether the defendant had failed to explore them.[69]

The relevance of these judgments for FHI lies not in their outcomes, but in their structure. Courts have treated the failure to explore alternatives, to adjust tactics, or to intervene in the face of foreseeable harm as legally significant omissions. While the doctrinal label remains command responsibility or precaution, the underlying logic is one of avoidability.

Intent-based frameworks trap prosecutors in a doctrinal cul-de-sac -unable to reach foreseeable harm without destabilising the distinction between targeting and proportionality. These developments suggest a path beyond that impasse: avoidability reasoning offers an alternative ground of inquiry that courts are already implicitly employing.

C. International Court of Justice

The ICJ has contributed indirectly. In Bosnia v Serbia, the Court treated foreseeability as relevant to breach of the duty to prevent genocide.[70] In its advisory opinions on Nuclear Weapons and Legal Consequences of the Construction of a Wall, the Court affirmed that humanitarian rules must be interpreted in light of contemporary conditions.[71] 

D. Domestic Judicial Review

Domestic courts have exhibited similar tendencies. The Israeli Supreme Court in the Targeted Killings case treated capture as a feasible alternative requiring consideration before lethal force - an analysis closely aligned with FHI’s avoidability inquiry.[72]

E. Convergence Without Consolidation

Across human rights law, international criminal law, and domestic review, courts have begun to treat foreseeability and preventability as legally relevant. What is missing is systematic jurisprudential consolidation. FHI offers a doctrinal anchor by identifying where avoidability fits within existing doctrine and what consequences follow from its breach.

VII. CONCLUSION: THE STAKES OF AVOIDABILITY

The argument for the Principle of Foreseeable Harm to Innocents is ultimately simple. Where civilians are visible and identifiable, where harm to them is predictable, and where alternatives exist that would spare their lives, the law must not permit attacks to proceed as though those facts were irrelevant.

The modern battlespace has transformed the moral landscape. Civilian presence is no longer hidden or unknowable; it is visible on sensors, tracked over time, and integrated into targeting decisions. Alternatives - delay, cancellation, different means and methods - are not speculative possibilities but real options that targeting cells consider and often reject. When harm nonetheless occurs, it cannot honestly be characterised as accidental or unforeseeable. It is the result of a choice.

FHI is ultimately animated by a simple proposition: knowability carries moral force. When a drone operator can see a child step into a courtyard, when thermal signatures identify a family in an upper floor, when CDE models predict with statistical precision that a strike will kill a father and daughter sitting at a kitchen table, the harm can no longer be treated as morally ‘incidental.’ What was once regrettably unforeseen becomes the result of conscious choice. Draper taught that modern law must refuse to ‘look away’ from the human realities of armed conflict; FHI places that refusal at the centre of its doctrinal structure.

FHI creates no new legal obligation; it makes explicit what AP I already requires when read in light of its object and purpose. In an operational environment where the battlespace is visible, harm is foreseeable, and alternatives exist, intention no longer carries the moral weight it once did. What FHI promises is intellectual honesty: ensuring deaths are not misclassified as unavoidable when they were anything but, and demanding that states bear the responsibility knowledge imposes.

FHI does not seek to revolutionise international humanitarian law. It seeks to complete it - giving operational effect to the precautionary structure that Article 57 already mandates, closing the gap that has allowed foreseeable and avoidable civilian deaths to be treated as lawful collateral damage. The avoidability gate is not new law; it is existing law rendered explicit for an operational environment the drafters could not fully anticipate.

The limitations of this framework must be acknowledged. FHI addresses cases where civilians are ascertainable, harm is foreseeable, and decision-makers have time to assess alternatives. This category - exemplified by the Mohammad Azam case - represents some of the most morally troubling instances of civilian death: harm that was visible, predicted, and preventable.[73] It does not exhaust civilian casualties. As Larry Lewis has documented, the majority of civilian harm arises where detection systems failed to identify civilian presence, intelligence was fragmentary, or the tempo of engagement foreclosed deliberation - circumstances that would not trigger the avoidability gate at all.[74] FHI addresses the cases it can address; a comprehensive approach requires action on multiple fronts.[75] 

A comprehensive approach to civilian protection demands action on multiple fronts: investment in protective capabilities, reform of targeting doctrine, strengthened accountability, and attention to systemic factors. FHI supplies one element of that architecture.

The stakes are not abstract. In every conflict where ISR makes civilian presence visible and precision weapons make discrimination possible, the question is whether those capabilities will generate corresponding obligations or remain legally inert. Whether this framework shapes practice depends on whether states, militaries, courts, and civil society are prepared to demand that foreseeability generate obligation, and that avoidability constrain permission.

If the law of armed conflict is to preserve any meaningful distinction between lawful warfare and unlawful killing, FHI’s core principle must be recognised:

Foreseeable and avoidable harm to innocents is impermissible harm.

Acknowledgments

I am grateful to Jens David Ohlin, Neta Crawford, Henry Shue, and Matthew Evangelista for their generous engagement with earlier drafts. Professor Crawford's comments on structure and the political dimensions of targeting capability were particularly influential. Professor Shue's work on capacity-indexed obligations informed the framework's core logic. Professor Evangelista's observations sharpened the emergency override's exclusionary scope. The Article builds on Professor Ohlin's foundational defence of the intent/foresight distinction; any errors in extending that framework are mine alone. I also thank Larry Lewis of the Center for Naval Analyses for detailed operational comments and his candid assessment of likely military legal objections. The views expressed, and any remaining errors, are solely my responsibility.

[1] See generally Larry Lewis, ‘Reducing and Mitigating Civilian Casualties: Enduring Lessons’ (CNA 2013); Human Rights Watch, ‘Precisely Wrong: Gaza Civilians Killed by Israeli Drone-Launched Missiles’ (2009); Azmat Khan and Anand Gopal, ‘The Uncounted’ New York Times Magazine (18 November 2017).

[2] See Neta C Crawford, Accountability for Killing: Moral Responsibility for Collateral Damage in America’s Post-9/11 Wars (Oxford University Press 2013) 1-15.

[3] The term ‘avoidability gate’ is mine; the underlying concept finds support in AP I Article 57(2)(a)(ii), which requires precautions ‘with a view to avoiding, and in any event to minimizing’ civilian harm.

[4] On the distinction between distinction and proportionality, see Jens David Ohlin, ‘Targeting and the Concept of Intent’ (2013) 35 Michigan Journal of International Law 79.

[5] Henry Shue and David Wippman, ‘Limiting Attacks on Dual-Use Facilities Performing Indispensable Civilian Functions’ (2002) 35 Cornell International Law Journal 559.

[6] See Crawford (n 2), particularly her analysis of how investment in precision capabilities transforms the moral and legal landscape of accountability for civilian harm.

[7] Chairman of the Joint Chiefs of Staff Instruction 3121.01B, Standing Rules of Engagement/Standing Rules for the Use of Force for U.S. Forces (13 June 2005).

[8] See G I A D Draper, The Red Cross Conventions (Stevens and Sons 1958); G I A D Draper, ‘Civilians and the NATO Status of Forces Agreement’ (1960) 23 Modern Law Review 553. Draper repeatedly emphasised restraint as the animating purpose of IHL rather than ex post justification.

[9] Ohlin (n 4).

[10] Crawford (n 2).

[11] Henry Shue, ‘Bombing to Rescue? NATO’s 1999 Bombing of Serbia’ in Deen K Chatterjee and Don E Scheid (eds), Ethics and Foreign Intervention (Cambridge University Press 2003) 97-117.

[12] John Mikhail, Elements of Moral Cognition: Rawls’ Linguistic Analogy and the Cognitive Science of Moral and Legal Judgment (Cambridge University Press 2011). Cognitive science research shows that humans possess intuitive moral knowledge distinguishing intended from merely foreseen harm - including recognition of the principle of double effect. Thus, FHI's distinction between avoidable and unavoidable harm is not an artificial legal construction but tracks deep features of how human moral cognition actually works. See also Stephen Roblin, ‘The Moral Public: Intent, Civilian Harm, and American Public Support for the Use of Force’ (PhD thesis, Cornell University 2025) (on file with the author). Roblin’s experimental research demonstrates that public support for military operations decreases as the intent behind civilian harm shifts from accidental to foreseeable to intentional, and that this effect is driven by moral intuition rather than legal knowledge.

[13] Peter von Hagenbach Case (1474), in Theodor Meron, Henry’s Wars and Shakespeare’s Laws (Oxford University Press 1993) 146-48.

[14] Instructions for the Government of Armies of the United States in the Field (Lieber Code), General Orders No 100 (1863) Articles 22 and 44.

[15] Hague Convention (IV) Respecting the Laws and Customs of War on Land (adopted 18 October 1907) 205 CTS 277 Articles 25-26.

[16] International Military Tribunal (Nuremberg), Judgment and Sentences (1 October 1946) 41 American Journal of International Law 172 (1947).

[17] United States v List (Hostage Case) (1948) 11 TWC 1230.

[18] In re Yamashita, 327 US 1 (1946). For analysis, see Allan A Ryan, Yamashita’s Ghost: War Crimes, MacArthur’s Justice, and Command Accountability (University Press of Kansas 2012).

[19] AP I Articles 48, 51 and 57.

[20] The term ‘capacity-indexed obligation’ is mine. The underlying concept - that the scope of the precautionary duty scales with the attacker’s capability - is implicit throughout the travaux préparatoires and the ICRC Commentary.

[21] Yves Sandoz, Christophe Swinarski and Bruno Zimmermann (eds), Commentary on the Additional Protocols of 8 June 1977 (ICRC/Martinus Nijhoff 1987) paras 2195-2200; Michael Bothe, Karl Josef Partsch and Waldemar A Solf, New Rules for Victims of Armed Conflicts (Martinus Nijhoff 1982) 359-61.

[22] Sandoz, Swinarski and Zimmermann (n 21) paras 2198-2200.

[23] Ohlin (n 4) 103-109, 111-114.

[24] ibid 124-126.

[25] Crawford (n 2) ch 4.

[26] ibid.

[27] ibid.

[28] ibid (emphasis in original).

[29] Shue (n 11) 97-117. Shue has argued that the stringency of the laws of war cannot be fixed independently of what compliance is reasonably possible in practice, and that changes in capacity bear directly on what moral and legal restraint may properly be demanded.

[30] Max Weber, ‘Politics as a Vocation’ in HH Gerth and C Wright Mills (eds), From Max Weber: Essays in Sociology (Oxford University Press 1946) 77-128; Mikhail (n 12); Roblin (n 12).

[31] Convention (II) with Respect to the Laws and Customs of War on Land (adopted 29 July 1899) 32 Stat 1803 preamble (Martens Clause); AP I Article 1(2).

[32] US Air Force, Air Force Doctrine Publication 3-70 Strategic Attack (25 July 2019) ch 4 (‘Elements of Effective Employment’). The doctrine identifies population-level psychological effects as intended strategic outcomes while simultaneously noting the law of war prohibits directly targeting civilians.

[33] AP I Article 51(5)(b).

[34] See Janina Dill, ‘The Definition of a Legitimate Target of Attack: Not More Than a Moral Plea?’ (2009) 103 Proceedings of the American Society of International Law 229; Michael A Newton, ‘Reframing the Proportionality Principle’ (2018) 51 Vanderbilt Journal of Transnational Law 867; Luigi Daniele, ‘The Concept of “Incidentality” in the Law of Targeting’ (2024) 29 JCSL 21.

[35] US Department of Defense, Law of War Manual (July 2023) para 5.6.5.

[36] Chairman of the Joint Chiefs of Staff Instruction 3121.01B (n 7).

[37] US forces in certain areas of operation in Afghanistan classified all military-age males in strike zones as combatants unless affirmatively identified as civilians - a reversal of the presumption required by AP I Article 50(1). See Jo Becker and Scott Shane, ‘Secret “Kill List” Proves a Test of Obama’s Principles and Will’ New York Times (29 May 2012); Conor Friedersdorf, ‘The Dangerous Precedent of Obama’s Kill List’ The Atlantic (2 May 2013). This practice was also documented in leaked intelligence assessments. See ‘The Drone Papers’ The Intercept (15 October 2015).

[38] Ohlin (n 4) 111–14; Prosecutor v Galić (Judgment) IT-98-29-T (5 December 2003) [58]–[59].

[39] Michael Walzer, Just and Unjust Wars: A Moral Argument with Historical Illustrations (5th edn Basic Books 2015) 155-56 (emphasis in original).

[40] Martin Shaw, The New Western Way of War: Risk-Transfer War and its Crisis in Iraq (Polity Press 2005).

[41] Larry Lewis, 'Hidden Negligence: Aug. 29 Drone Strike is Just the Tip of the Iceberg' (Just Security, 16 December 2021). Lewis, analysing approximately two thousand civilian harm incidents as a US advisor (2007-2021), observes that 'the great majority of cases are not where the military detects collateral concerns and goes ahead anyway, but rather they fail to detect those risks.'

[42] The requirement of 'constant care' in AP I Article 57(1) has been interpreted by the ICRC to include affirmative duties of information-gathering. See Jean-Marie Henckaerts and Louise Doswald-Beck (eds), Customary International Humanitarian Law (CUP 2005) vol I, Rule 15 ('In the conduct of military operations, constant care must be taken to spare the civilian population, civilians and civilian objects'). The US DoD Law of War Manual acknowledges that 'Commanders and other decision-makers must make a good faith assessment... with the information reasonably available at the time' but notes that this includes the 'obligation to seek available information': US Department of Defense, Law of War Manual (June 2015, updated December 2023) § 5.2.3.2.

[43] See generally Center for Civilians in Conflict and Stimson Center, Tracking Implementation of the Civilian Harm Mitigation and Response Action Plan (CHMR-AP) (November 2024); Crawford (n 2) ch 9.

[44] The DoD Law of War Manual acknowledges that attacks may proceed where civilian harm might be avoidable if commanders had certain assets but operational constraints dictate otherwise: DoD Law of War Manual (n 35) § 5.10.1.1. However, it also recognises that feasibility 'is not a justification for failing to take any precautions' and depends on 'all circumstances ruling at the time, including humanitarian and military considerations': ibid § 5.10.1.

[45] See generally Larry Lewis and Sarah Holewinski, ‘Changing of the Guard: Civilian Protection for an Evolving Military’ (2013) 2 PRISM 57.

[46] US Air Force, Air Force Doctrine Publication 3-60 Targeting (12 November 2021); NATO Standardization Office, ‘NATO Standard, AJP-3.9, Allied Joint Doctrine for Joint Targeting, Edition B, version 1’ (November 2021); Council of the European Union, ‘Avoiding and Minimizing Collateral Damage in EU-led Military Operations Concept’ EEAS (2015)772 Rev 8 (3 February 2016); Ministry of Defense, Republic of France, ‘Collateral Damage Estimation’ Joint Publication (FRA) JP-3.9.9, CDE (2014).

[47] For discussion of NCV thresholds and their fluctuation, see Craig Jones, The War Lawyers: The United States, Israel, and Juridical Warfare (Oxford University Press 2020).

[48] US Department of Defense, Civilian Harm Mitigation and Response Action Plan (25 August 2022) 1 ('Nothing in this plan is intended to suggest that existing DoD policies or practices are legally deficient or that the actions to be implemented... are legally required, including under the law of war').

[49] The policy/law distinction has enforcement consequences. The Obama administration's 'near-certainty' standard for strikes outside active hostilities was adopted as policy and subsequently revoked, whereas Article 57's requirements are treaty-based. See Presidential Policy Guidance (22 May 2013); Charlie Savage, 'Trump Revokes Obama-Era Rule on Reporting Drone Strike Deaths' (New York Times, 6 March 2019).

[50] For documented resistance to capability-indexed precautionary standards, see Kenneth Watkin, 'Assessing Proportionality: Moral Complexity and Legal Rules' (2005) 8 Yearbook of International Humanitarian Law 3, 42-47; The US position on Article 57 reservations reflects similar concerns: see Adam Roberts and Richard Guelff (eds), Documents on the Laws of War (3rd edn, OUP 2000) 509-12.

[51] ICRC Commentary (n 21) para 2198 ('The word "feasible" means practicable or practically possible. Taking into account all circumstances ruling at the time, including humanitarian and military considerations, precautions should be taken which are "feasible," i.e., which can actually be carried out'). The Commentary explicitly links feasibility to 'the technical means of identification available': ibid para 2195.

[52] This objection inverts the moral logic of distinction. As Walzer argues, the capacity to discriminate is precisely what distinguishes lawful warfare from indiscriminate violence: Michael Walzer, Just and Unjust Wars (4th edn, Basic Books 2006) 152-56. See also Shue (n 11) 113 ('A state-of-the-art military that has the capacity to discriminate... cannot coherently claim that it lacks the obligation to use that capacity').

[53] Compare the structure of avoidability review with proportionality's reliance on 'anticipated' military advantage, which courts have found difficult to assess ex post without impermissible hindsight reasoning. See Prosecutor v Galić (Judgment) IT-98-29-T (5 December 2003) [58] (noting difficulties in reconstructing the commander's assessment); UK Ministry of Defence, The Manual of the Law of Armed Conflict (OUP 2004) para 5.33.2 (acknowledging that proportionality assessment 'must be made in good faith and in light of the circumstances prevailing at the time').

[54] The indeterminacies of proportionality have been extensively documented. See Adil Ahmad Haque, Law and Morality at War (OUP 2017) 188-206; Michael Newton, 'Reconsidering Reprisals' (2010) 20 Duke Journal of Comparative and International Law 361, 378-82; Dill (n 34) 143-52. FHI's contribution lies not in resolving these indeterminacies but in confining proportionality to cases where they must be confronted.

[55] Walzer (n 39) 251-68.

[56] Henry Shue, ‘Liberalism: The Impossibility of Justifying Weapons of Mass Destruction’ in Sohail H Hashmi and Steven P Lee (eds), Ethics and Weapons of Mass Destruction: Religious and Secular Perspectives (Cambridge University Press 2004) 139-62.

[57] For comprehensive treatment of human shields in IHL, see Michael N Schmitt, ‘Human Shields in International Humanitarian Law’ (2009) 47 Columbia Journal of Transnational Law 292; Stéphane Bouchié de Belle, ‘Chained to Cannons or Wearing Targets on Their T-Shirts: Human Shields in International Humanitarian Law’ (2008) 90 International Review of the Red Cross 883.

[58] AP I Article 51(7). The prohibition is recognised as customary international law applicable in both international and non-international armed conflicts. Jean-Marie Henckaerts and Louise Doswald-Beck, Customary International Humanitarian Law (ICRC 2005) rule 97.

[59] Rome Statute of the International Criminal Court, Article 8(2)(b)(xxiii), (adopted 17 July 1998, 2187 UNTS 90).

[60] For analysis of operational alternatives in human-shielding scenarios, see Michael N Schmitt, 'Human Shields in International Humanitarian Law' (2009) 47 Columbia Journal of Transnational Law 292, 330-38; Janina Dill, Legitimate Targets? Social Construction, International Law, and US Bombing (CUP 2014) 175-82. For discussion of non-kinetic alternatives including cyber operations and electronic warfare, see Michael N Schmitt (ed), Tallinn Manual 2.0 on the International Law Applicable to Cyber Operations (CUP 2017) Rules 113-15; Giacomo Biggio, 'Regulating Non-Kinetic Effects of Cyber Operations' (2025) 30 JCSL 241.

[61] The voluntary/involuntary distinction derives from AP I Article 51(7) ('The presence or movements of the civilian population or individual civilians shall not be used to render certain points or areas immune from military operations, in particular in attempts to shield military objectives from attacks'). For critique of the binary framework and analysis of the spectrum of coercion, see Neve Gordon and Nicola Perugini, 'The Politics of Human Shielding: On the Resignification of Space and the Constitution of Civilians as Shields in Liberal Wars' (2016) 34 Environment and Planning D: Society and Space 168, 172-78.

[62] This formulation draws on Matthew Evangelista’s observation regarding the inherent difficulty of addressing situations where both parties may be violating their obligations.

[63] The European Court of Human Rights has repeatedly required states to demonstrate that lethal force was ‘absolutely necessary’ and that adequate precautions were taken. See McCann v United Kingdom (1995) 21 EHRR 97; Nachova v Bulgaria (2005) 42 EHRR 43 [148]-[149].

[64] McCann v United Kingdom (1995) 21 EHRR 97.

[65] Finogenov v Russia (2011) App nos 18299/03 and 27311/03 (ECtHR) [208]; Tagayeva v Russia, App nos 26562/07 and others (ECtHR, 13 April 2017).

[66] Isayeva v Russia (2005) App no 57950/00 (ECtHR) [189]-[191].

[67] Prosecutor v Galić (Judgment) IT-98-29-T (5 December 2003) [58]-[59].

[68] Prosecutor v Kupreškić (Judgment) IT-95-16-T (14 January 2000).

[69] Prosecutor v Ntaganda (Judgment) ICC-01/04-02/06 (8 July 2019) [916]-[918].

[70] Application of the Convention on the Prevention and Punishment of the Crime of Genocide (Bosnia and Herzegovina v Serbia and Montenegro) (Judgment) [2007] ICJ Rep 43.

[71] Legality of the Threat or Use of Nuclear Weapons (Advisory Opinion) [1996] ICJ Rep 226; Legal Consequences of the Construction of a Wall in the Occupied Palestinian Territory (Advisory Opinion) [2004] ICJ Rep 136.

[72] Public Committee Against Torture in Israel v Government of Israel (Targeted Killings) HCJ 769/02 (11 December 2006) [40]-[46].

[73] See Part I above (Mullah Mansour strike). For a typology distinguishing 'genuine accidents' from 'systemic collateral damage' and 'double effect killing,' see Crawford (n 2) 21-26.

[74] Larry Lewis (n 41).

[75] See Crawford (n 2) ch 9; CIVIC and Stimson Center (n 43) 12-18. FHI addresses one dimension of a multi-dimensional problem.