A Theory of Jus in Bello Proportionality

Adil Ahmad Haque

in  WEIGHING  LIVES:  COMBATANTS  &  CIVILIANS  IN  WAR  (Jens  David  Ohlin,  Larry  May,  
Claire  Finkelstein  eds.,  forthcoming  2016).    

The proportionality principle 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’.1 The importance of this
prohibition is difficult to overstate. While the precautions principle regulates
how armed forces may pursue a particular military advantage, the
proportionality principle regulates whether a particular military advantage
may be pursued or must be abandoned. Even if attacking forces select
weapons, tactics, and targets that best avoid or most reduce harm to
civilians, even at their own risk, they must forego one path to victory if the
expected civilian losses are too great. Rather than inflict disproportionate
harm on civilians, attacking forces must find another way to win.
An account of jus in bello proportionality must satisfy two apparently
conflicting demands. First, such an account must explain how we can
rationally compare civilian losses with military advantages. At the same
time, such an account must apply symmetrically to all parties to every
conflict independently of the jus ad bellum status of their war effort.
Existing accounts of jus in bello proportionality satisfy either one demand or
the other. In this chapter, I offer a new account that satisfies both demands.
In addition, I offer an account of how combatants should cope with predictive
uncertainty regarding the likely humanitarian costs and likely military
benefits of their actions as well as moral uncertainty regarding their relative
weight.
I argue that an attack that inflicts incidental harm on civilians is
objectively proportionate only if it prevents substantially greater harm to the
attacking force or to its civilians over the remainder of the conflict. This
account reflects the moral asymmetry between doing harm and allowing
harm while looking beyond particular tactical engagements to the broader
operational context. I argue that an attack is epistemically proportionate
only if the attacker reasonably believes — on the basis of decisive epistemic
reasons — that the attack will prove objectively proportionate. Put another
way, an attacker must reasonably believe that, of all the possible outcomes of
the attack, it is probable that the outcome of the attack will be objectively
proportionate.

1 Protocol Additional to the Geneva Conventions of 12 August 1949, and relating to

the Protection of Victims of International Armed Conflicts, 8 June 1977 (Protocol I) art 51(5).
See also International Committee of the Red Cross, Customary International Humanitarian
Law, vol 1 (CUP 2009) 46.

Electroniccopy
Electronic copy available
available at:
at:https://ssrn.com/abstract=2575841
http://ssrn.com/abstract=2575841
 My account of jus in bello proportionality is in one way more
determinate than existing accounts — for example, it does not compare
incommensurable values but instead compares immediate losses to civilians
and future losses to civilians and to attacking forces. At the same time, my
account must still grapple with the predictive uncertainty inherent in
determining whether immediate losses inflicted will be redeemed by future
losses avoided. Accordingly, I explore a number of decision procedures and
rules of engagement that officers may use to make the best possible decision
given the limited information available to them.

Elements of the Proportionality Rule

Let us begin by briefly reviewing key elements of the proportionality rule.

First, civilian losses are excessive in relation to military advantage just in
case the former exceed, outweigh, or are unjustified by the latter.2 Strangely,
it is sometimes suggested that civilian losses are ‘excessive’ in relation to
military advantage only if the former outweigh the latter by a substantial
margin.3 This suggestion has no obvious basis in language or logic. In
ordinary language, ‘excessive’ simply means ‘exceeding’ or ‘going beyond’
some normative standard.4 Moreover, it seems bizarre to concede that the
humanitarian considerations against an attack outweigh the military
considerations in favor of the attack yet insist that the attack is nevertheless
justified.
Second, international law prohibits attacks which may be expected to
cause civilian losses which would be excessive in relation to the military
advantage anticipated. In other words, the lawfulness of an attack depends
not on its actual consequences but on its reasonably foreseeable
consequences.5 Attackers must do everything feasible to assess whether an

2 See US Dep’t of Defense, Law of War Manual 2.4.1.2 (‘Proportionality generally

weighs the justification for acting against the expected harms to determine whether the
latter are disproportionate in comparison to the former’).
3 See, eg, Israel Ministry of Foreign Affairs, The 2014 Gaza Conflict: Factual and

Legal Aspects (2015) 185 (‘As long as there is no significant imbalance between the expected
collateral damage and the anticipated military advantage, no excessiveness exists’); Geoffrey
S Corn & Gary P Corn, ‘The Law of Operational Targeting: Viewing the LOAC Through an
Operational Lens’ (2012) 47 Tex. Int'l L.J. 337, 365 (writing that ‘an attack does not become
[unlawful] when the collateral damage or incidental injury is slightly greater than the
military advantage anticipated (what is suggested by the term ‘disproportionate’) but only
when those effects are excessive’).
4 See, eg, Merriam-Webster Dictionary (defining ‘excessive’ as ‘exceeding what is

usual, proper, necessary, or normal’); Oxford English Dictionary (‘More than is necessary,
normal, or desirable’).
5 Cf Commentary on the HPCR Manual on International Law Applicable to Air and

Missile Warfare 91 (‘The standard is objective in that expectations must be reasonable’).

Electroniccopy
Electronic copy available
available at:
at:https://ssrn.com/abstract=2575841
http://ssrn.com/abstract=2575841
attack will cause excessive civilian losses and to cancel or suspend an attack
if it becomes apparent that it will do so.6 It follows that

In determining whether an attack was proportionate it is necessary to

examine whether a reasonably well-informed person in the
circumstances of the actual perpetrator, making reasonable use of the
information available to him or her, could have expected excessive
civilian casualties to result from the attack.7

Put another way, international law does not prohibit objectively

disproportionate attacks but instead prohibits epistemically disproportionate
attacks. Much more on this below.
Third, only military advantages that are concrete and direct can legally
justify civilian losses. A military advantage is ‘any consequence of an attack
which directly enhances friendly military operations or hinders those of the
enemy’ such as disabling opposing combatants, destroying their equipment,
denying them opportunities to attack, and creating opportunities to attack
them.8 By contrast, ‘forcing a change in the negotiating position of the enemy
only by affecting civilian morale does not qualify as military advantage’.9 A
military advantage is concrete only if it is substantial and clearly identifiable.
A military advantage is direct only if it is proximately caused, either without
further intervening agency (as with the destruction of weapons to prevent
their future use) or with reasonably foreseeable intervening agency (as with
strikes intended to lead an adversary to divert troops or resources away from
one’s true objective).
It has been claimed that military advantages are concrete and direct
only if they are ‘relatively close’ to the attack in space and time, and that
advantages ‘which would only appear in the long term should be
disregarded’.10 However, it seems clear that the destruction of weapons in a
factory or armory would yield a concrete and direct advantage even if the
weapons would not have been sent to the front lines for many weeks.
Accordingly, a concrete and direct advantage ‘may or may not be temporally
or geographically related to the object of the attack’ so long as it is
‘foreseeable by the [attacker] at the relevant time’.11
When the achievement of a military advantage requires coordinated
attacks, the proportionality rule compares the harm expected and the

6 Protocol I arts 57(2)(a)(iii) & 57(2)(b); ICRC, Customary IHL Study 58 & 60.
7 ICTY, Prosecutor v. Galić, Judgment, IT-98-29, Trial Chamber, December 5, 2003,
para 58.
8 Commentary on the HPCR Manual 45.
9 Ibid.
10 ICRC, Protocol I Commentary para 2209.
11 See, eg, Elements of Crimes n. 36.

Electronic copy available at: https://ssrn.com/abstract=2575841

military advantage anticipated from the operation as a whole.12
Paradigmatically, it may be necessary to destroy several bridges in order to
prevent the movement of troops or supplies. The military advantage of
destroying any one bridge may be trivial if the opposing force could simply
use one of the other bridges. Accordingly, we should compare the advantage
anticipated and the harm expected from destroying all the bridges.
Fourth, it is sometimes claimed that ‘Remote harms resulting from
[an] attack do not need to be considered in a proportionality analysis’.13
However, it is hard to believe that temporally remote harms foreseeably
resulting from placing a mine or improvised explosive device in a civilian
area do not need to be considered in a proportionality analysis. Similarly,
destroying bridges or roads necessary to bring food or medicine to a civilian
population may predictably result in civilian deaths when existing supplies
run out.14 It seems that such foreseeable remote harms also need to be
considered in a proportionality analysis.
Fifth, an attack may be rendered disproportionate by loss of civilian
life, injury to civilians, or damage to civilian property.15 Other bad
consequences of an attack may render the attack morally impermissible all
things considered, but will not render an attack legally disproportionate.
My account of jus in bello proportionality compares harms inflicted
with harms prevented, and is compatible with different accounts of how to
compare death, injury, and property damage to one person or across persons.
To take but one example, nothing in my account precludes measuring the

12 See, eg, UK Ministry of Defense, Law of Armed Conflict Manual (OUP 2005)
5.33.5; ICRC, Protocol I Commentary para 2218.
13 US Dep’t of Defense, Law of War Manual 5.12.3.
14 See ICTY, Prosecutor v Prlic, Judgment, IT-04-74, Trial Chamber, May 29, 2013,

paras 1582-84.
15 Environmental damage raises interesting issues. For example, Protocol I prohibits

methods or means of warfare which may be expected to cause widespread, long-term and
severe damage to the natural environment. Protocol I, art 55(1). Accordingly, environmental
damage is prohibited if it is severe even if it is not excessive, but is not prohibited if it is
excessive but not severe.
In contrast, the Rome Statute of the International Criminal Court recognizes a war
crime of knowingly inflicting ‘widespread, long-term and severe damage to the natural
environment which would be clearly excessive in relation to the concrete and direct overall
military advantage anticipated’. Rome Statute, art 8(2)(b)(iv). Accordingly, environmental
damage is punishable only if it is both severe and clearly excessive.
Finally, the ICRC takes the view that customary international law prohibits attacks
‘which may be expected to cause incidental damage to the environment which would be
excessive in relation to the concrete and direct military advantage anticipated’. ICRC,
Customary IHL Study 143. Accordingly, environmental damage is prohibited if it is
excessive even if it is not severe. I will not attempt to reconcile these approaches here.
See also Eliav Lieblich, ‘Beyond Life and Limb: Exploring Incidental Mental Harm
under International Humanitarian Law’, in Derek Jinks, Jackson Nyamuya Maogoto &
Solon Solomon (eds), Applying International Humanitarian Law in Judicial and Quasi-
Judicial Bodies: International and Domestic Aspects (TMC Asser 2014) 185.

Electronic copy available at: https://ssrn.com/abstract=2575841

harmfulness of death, injury, and property damage in terms of quality-
adjusted life years or ‘QALYs’ and simply aggregating such QALYs across
persons. Although my account prominently includes a deontic asymmetry
between doing and allowing harm, it is compatible with broadly utilitarian
accounts of measuring and aggregating harms.
As it happens, I do not believe that it would be permissible to
collaterally kill one innocent person as a side-effect of preventing small
harms to many people, even if the total harm prevented (however measured)
would be substantially greater than the total harm inflicted. On this roughly
prioritarian view, we should consider not only total harm inflicted or
prevented but also distributions of harm that leave some much worse off than
others.16
Accordingly, it may be impermissible to collaterally kill or injure
civilians merely to prevent a great deal of damage to military or civilian
property. Conversely, it may be permissible to collaterally damage a great
deal of civilian property if necessary to prevent soldiers or civilians from
being killed or injured. In principle, those who would be killed or injured
would be made far worse off than those who would lose only their property.
Of course, in practice, damage to military or civilian property often leads to
death, injury, destitution, or illness. Accordingly, proportionality may
depend not on property damage as such but instead on its predictable
downstream consequences.
Similarly, it may be objectively impermissible to kill civilians to
prevent many more civilians or soldiers from being injured. Conversely, it
may be objectively permissible to prevent soldiers or civilians from being
killed even if it means collaterally injuring many more civilians. In principle,
those who would be killed would be made far worse off than those who would
be injured. Of course, in practice, it is often very hard to predict whether an
act will kill or injure. Accordingly, often it will be epistemically permissible
to launch attacks that will (in fact) collaterally kill civilians in order to
prevent attacks that would (in fact) have only injured. In any event, my
account of jus in bello proportionality does not depend on any particular
resolution of these difficult issues of measurement and comparison.
Finally, civilian losses may be legally justified only by advantages that
that are causally downstream from harming, capturing, or neutralizing
opposing combatants or military objectives. In contrast, the proportionality
principle excludes advantages that are causally downstream from collateral
harm to civilians. As we saw in the introduction, attackers may not
opportunistically take advantage of the presence of civilians to obtain
military advantages. Accordingly, the fact that collateral harm to civilians
would spread terror or damage morale may not render an otherwise
disproportionate attack proportionate.

16 For more on prioritarianism, see Derek Parfit, ‘Equality and priority’ (1997) 10

Ratio 202–21.

Electronic copy available at: https://ssrn.com/abstract=2575841

Incommensurable and Incomparable Values

On its face, jus in bello proportionality seems to call for a comparison

between two incommensurable values — civilian losses and military
advantage — irreducible to any more basic value.17 Civilian losses may be
ranked according to their relative moral gravity, military advantages
according to their relative contribution to military victory, but there appears
to be no single standard according to which both may be ranked by their
relative value.
Although civilian losses and military advantages seem
incommensurable, they are not entirely incomparable.18 Certainly, we often
make confident proportionality judgments in extreme cases. As Henry Shue
observes, ‘[i]t is undeniably excessive to inflict very large civilian losses for
the sake of a very small military advantage’.19
For example, the ICRC writes that ‘the presence of a soldier on leave
obviously cannot justify the destruction of a village. Conversely, if the
destruction of a bridge is of paramount importance for the occupation or non-
occupation of a strategic zone, it is understood that some houses may be hit,
but not that a whole urban area be leveled’.20 Similarly, the Israeli High
Court writes that ‘shooting at [a sniper firing on soldiers or civilians] is
proportional even if as a result, an innocent civilian neighbor or passerby is
harmed. That is not the case if the building is bombed from the air and
scores of its residents and passersby are harmed’.21 However, our less
confident judgments in close and intermediate cases suggest that military
advantages and civilian losses are at most imprecisely or roughly comparable.
If civilian losses and military advantages are only imprecisely
comparable, we should not expect jus in bello proportionality to provide clear

17 See, eg, APV Rogers, Law on the Battlefield (1996) 17 (‘Some delegations at the
diplomatic conference at which Protocol I was negotiated ... were reluctant to include any
reference to the proportionality rule because of the difficulty of comparing things that were
not comparable (i.e. military advantage and civilian losses)’); Michael N Schmitt, ‘The
Principle of Discrimination in 21st Century Warfare’ (1999) 2 Yale Hum. Rts. & Dev. L.J. 143,
151 (‘Optimally, balancing tests compare like values. However, proportionality calculations
are heterogeneous, because dissimilar value genres — military and humanitarian — are
being weighed against each other. How, for example, does one objectively calculate the
relative weight of an aircraft, tank, ship, or vantage point in terms of human casualties?’).
18 See Ruth Chang, ‘Introduction’ in Ruth Chang (ed), Incommensurability,

Incomparability, and Practical Reason (1997) 14 (noting that incommensurable values often
permit ‘nominal-notable’ comparisons in extreme cases).
19 Henry Shue, ‘Proportionality in War’ in Gordon Martel (ed), The Encyclopedia of

War (2012) 6.
20 ICRC, Protocol I Commentary para 2214.
21 HCJ 769/02, Pub. Comm. Against Torture in Israel v. Gov't of Israel, Judgment,

Dec. 11, 2005, para 46. See also HCJ 2056/04, Beit Sourik Village Council v. Israel,
Judgment, June 30, 2004, para 41.

Electronic copy available at: https://ssrn.com/abstract=2575841

guidance in the vast majority of tactical situations that attacking forces
confront. Nevertheless, Shue proposes that we can improve our
proportionality judgments by sorting particular military advantages and
civilian losses into rough categories along the following lines:

Military Advantage Civilian Losses

Level 1 Important Moderate
Level 2 Compelling Severe
Level 3 Decisive Catastrophic

According to Shue, ‘it is excessive to inflict civilian losses of a category higher

than the category of military advantage anticipated’.22 On this view, an
attacking force may inflict moderate civilian losses in pursuit of an
important, compelling, or decisive military advantage; severe civilians losses
in pursuit of a compelling or decisive advantage; and catastrophic civilian
losses only in pursuit of a decisive military advantage.
Shue does not attempt to fix the boundaries of each category, but he
imagines that the task of doing so would proceed along parallel tracks, with
military experts categorizing military advantages according to military
standards and ‘morally sensitive’ individuals categorizing civilian losses
according to moral standards. Independent moral judgment would then be
exercised to sort the categories created by the two groups into three (or more)
levels. Each level would then contain categories of losses and advantages
that are roughly equal in moral weight.
The virtue of Shue’s approach is that it can apply symmetrically to all
sides of a conflict independently of the justice of their respective war aims.
The basic shortcoming of Shue’s approach is that it leaves our intuitive
judgments opaque. Since it does not identify the moral principles or
empirical assumptions underlying our judgments, it cannot tell us when
those principles are inapplicable or those assumptions are unsound. Since it
does not illuminate the basis of our intuitive proportionality judgments in
extreme cases, it cannot help us make inferential proportionality judgments
in non-extreme cases. This approach cannot tell us what to consider or ignore
when we make such judgments, or what circumstances might render our
intuitive judgments more or less reliable.
On a practical level, Shue’s approach organizes our intuitive moral
judgments but cannot replace or improve upon them. Certainly, we can rank
civilian losses by their moral gravity and military advantages by their
military utility, but we lack a common standard of value through which the
two rankings can be integrated. True, we could directly judge one item in the
first ranking comparable to one item in the second ranking, and use this

22 Shue, ‘Proportionality in War’ 7. For a similar approach, see Jason D. Wright,

‘‘Excessive’ ambiguity: analysing and refining the proportionality standard’ (2012) 94

International Review of the Red Cross 819, 852.

Electronic copy available at: https://ssrn.com/abstract=2575841

direct comparison as an anchor for integrating the remainder of the two
rankings. However, simply ranking two items does not reveal the degree of
difference between them. It is not enough to know that a ‘compelling’
military advantage is greater than an ‘important’ military advantage and
that ‘severe’ civilian losses are greater than ‘moderate’ civilian losses. We
also need to know how much greater advantages or losses in one category are
than advantages or losses in another category. Only then can we use our
anchor point to generate parallel rankings of imprecisely comparable values.
It seems unlikely that Shue’s approach could generate such a complete
ranking.

Intrinsic and Instrumental Values

Thomas Hurka and Jeff McMahan reject the view that civilian losses and
military advantages are incommensurable values on the grounds that
military advantages have no intrinsic value at all.23 Indeed, soldiers ought
not harm opposing combatants, capture strategic territory, or destroy
military equipment for its own sake. Such military advantages have
instrumental or derivative value only if they contribute to some further,
intrinsically valuable state of affairs. Importantly, the defeat of an opposing
armed force has intrinsic value only if one fights for a just cause — such as
national self-defense or humanitarian intervention — that morally justifies
resorting to or continuing the use of military force.
Hurka and McMahan conclude that the value of a military advantage,
if any, lies in the contribution that it makes to the achievement of a just
cause. Conversely, a military advantage that contributes to an unjust cause
has no moral value. Hurka and McMahan therefore reject the independence
of jus ad bellum just cause and jus in bello proportionality and with it the
symmetrical application of jus in bello proportionality. Put another way, jus
in bello proportionality is just a special application of jus ad bellum
proportionality. Just as jus ad bellum proportionality compares civilian
losses inflicted by the war as a whole with the importance of a just cause, jus
in bello proportionality compares civilian losses inflicted by a particular
military operation with the contribution of that operation to the achievement
of the same just cause.
For just combatants, this moral standard makes for an impossible
decision procedure. Soldiers would first have to measure the moral
importance of their war aims, since a similar contribution to a more
important war aim would justify more extensive civilian losses. Soldiers

23 See Thomas Hurka, ‘Proportionality in the Morality of War’, Philosophy & Public

Affairs 33 (2005) 34; Jeff McMahan, ‘Proportionality and Necessity in Jus in Bello’, in Helen
Frowe and Seth Lazar (eds), The Oxford Handbook of the Ethics of War (OUP 2016);
McMahan, ‘Proportionate Defense’ (2013-2014) 23 Journal of Transnational Law and Policy 1,
20-21.

Electronic copy available at: https://ssrn.com/abstract=2575841

would then have to measure the degree to which the achievement of a
particular military advantage would increase the probability of achieving
their war aims, discount the value of their war aims by this marginal
increase in probability, and compare the resulting expected value with the
civilian losses that they expect to inflict. Finally, if their just cause depends
on collective values such as national self-determination then soldiers will
have to somehow compare incommensurable (individual and collective)
values. Nor is it clear how a more manageable decision procedure could be
derived from the moral standard proposed.
Importantly, on this view, unjust combatants cannot conform to jus in
bello proportionality. A military advantage cannot inherit moral value from
a war aim that has no moral value to pass on, and a military advantage with
no moral value cannot morally justify inflicting civilian losses. On this view,
every harm that unjust combatants inflict on civilians is morally
disproportionate. It follows that, on this view, proportionality prohibits but
cannot regulate the conduct of unjust combatants. This view provides no
moral guidance to combatants who are forced to fight without a just cause.
To be sure, unjust combatants can still choose to minimize the harm that
they inflict on civilians in pursuit of their war aims and even to place
themselves at greater risk to reduce the risks that they impose on civilians.
But since unjust combatants cannot pursue their unjust war aims by
proportionate means it is useless for them to try. Nor can observers judge a
particular military operation proportionate or disproportionate simply by
comparing the military advantage that it aims to achieve with the civilian
losses that it predictably inflicts. Except in extreme cases, observers must
first judge the war as a whole. If this view prevails then proportionality
judgments will become even more politicized than they are already.
Hurka and McMahan offer an account of jus in bello proportionality
that is morally intelligible but applies asymmetrically and yields an
impractical decision procedure. Shue offers an account that applies
symmetrically but is morally inexplicable and of limited practical use. In the
following sections I will try to do better.

Inflicting and Preventing Harm

As we saw in the introduction, killing and injuring other human beings

presumptively infringes their basic moral rights. In some cases, these others
may, by their voluntary conduct, make themselves morally liable to be killed
or injured. In such cases, the relevant moral rights are forfeited and not
infringed. Alternatively, killing or injuring some may be the lesser of two
evils, necessary to prevent greater harm to others. More precisely, killing or
injuring some is morally permissible as a necessary means of preventing far
greater harm to others or as an unavoidable side-effect of preventing
substantially greater harm to others. This qualification reflects the view

Electronic copy available at: https://ssrn.com/abstract=2575841

that, other things equal, intentionally doing harm is morally much worse —
that is, harder to justify — than unintentionally doing harm which is, in
turn, morally substantially worse than allowing harm. In such cases, the
relevant moral rights are overridden and therefore justifiably infringed.
I propose that it is jus in bello proportionate to collaterally kill or harm
civilians as an unintended side-effect of achieving a military advantage that
will prevent or enable one to prevent substantially greater harm to one’s own
forces or civilians as one pursues military victory over the remainder of the
conflict. The last qualification (‘over the remainder of the conflict’) is
important because, while attacks on opposing forces sometimes prevent
imminent harm to attacking forces or to civilians, most attacks on military
equipment or strategic locations prevent or avoid such harm only over the
medium- or long-term.
If the proposed standard is satisfied then the moral rights of the
civilians harmed are permissibly overridden to protect the moral rights of
others. On this view, the moral value of a military advantage lies in the
harm to soldiers and civilians that it prevents. This standard applies
symmetrically to all sides of a conflict, independently of their war aims, yet
identifies a morally compelling explanation for when military advantage
justifies civilian losses.
While my account may seem unfamiliar to readers steeped in military
practice, it should not. Targeting decisions are often driven by considering
the criticality of a target, including the target’s importance to the adversary’s
ability to conduct operations (value), the time interval between a strike on
the target and a measurable impact on the adversary’s ability to conduct
operations (depth), and the time and cost required for the adversary to regain
its functional capability (recuperation).24 My account simply focuses
targeting decisions on the adversary’s capacity to inflict serious harm.
Let me illustrate the account by applying it to a series of skeletal
cases. To keep things simple, these cases will assume an international armed
conflict between state armed forces. At the end of the chapter I will apply the
account to targeted killing operations within a non-international armed
conflict between state armed forces and organized armed groups.
The first case is the simplest:
Simple Prevention: State A is at war with State B. If State A destroys
State B’s long-range missiles then some of State B's civilians will be
killed as a side-effect. However, if State A does not destroy State B's
long-range missiles then many more of State A’s soldiers and civilians
will be killed by those missiles over the remainder of the conflict.
A strike on the long-range missiles directly prevents State B from harming
State A’s soldiers and civilians (hence Simple Prevention). On my view, this
operation seems proportionate since the immediate harm that it inflicts is

24 US Joint Chiefs of Staff, Joint Publication 3-60: Joint Targeting (2007) D-2–D-4.

Electronic copy available at: https://ssrn.com/abstract=2575841

substantially less than the future harm that it prevents. Importantly, we can
judge the operation without reference to the war aims of either State A or
State B. More on this below.
It might seem that the preceding case calls for a simple comparison
between the consequences of action and the consequences of inaction but that
is not quite correct. The harm that an attack will prevent depends not only
on what the opposing party will do if the attack is not carried out but also on
what the attacking party will do if the attack is not carried out and how the
opposing force will respond. The preceding scenario assumes that if the
operation is not launched then State A will not simply surrender but will
continue to fight State B by other means. However, State A’s alternative
means of fighting will not prevent State B from launching its long-range
missiles. In this sense, my view assumes that parties will pursue victory
until they either win or lose. My view compares only the harms inflicted and
prevented by intervening military operations.
Now let us add some facts to the first case:
Triple Prevention: If State A destroys State B's anti-aircraft missiles
then some of State B's civilians will be killed as a side-effect. However,
if State A does not first destroy State B's anti-aircraft missiles then
State A’s air forces will be shot down before they can destroy State B’s
long-range missiles.
A strike on the anti-aircraft missiles will not directly prevent State B from
harming State A’s soldiers and civilians. Instead, the strike will prevent
State B from preventing State A from preventing State B from harming State
A’s soldiers and civilians (hence Triple Prevention).
On my view, jus in bello proportionality is satisfied only if the losses
inflicted on State B’s civilians by both the strike on the anti-aircraft missiles
and the subsequent strike on the long-range missiles are substantially less
than the losses that the long-range missiles would inflict on State A’s soldiers
and civilians. Cases like Triple Prevention illustrate the general truth that
the proportionality of an attack depends not only on the harm that it
prevents but also on the harm that it enables additional attacks to prevent.
For example, ‘If, in order to prevent the enemy’s army from advancing,
planners decide to destroy all the bridges that span a river, . . . each driver or
pilot may judge that his own action is disproportionate, [but] the operation as
a whole may meet the proportionality requirement’.25
Now let us change the facts of the second case slightly:
Timely Prevention: If State A destroys State B's anti-aircraft missiles
then some of State B's civilians will be killed as a side-effect. However,
if State A does not first destroy State B's anti-aircraft missiles then it
will take longer for State A’s air forces to destroy State B’s long-range

25 Jean-François Quéguiner, ‘Precautions under the law governing the conduct of

hostilities’, Int’l Rev. Red Cross 88 (2006) 804.

Electronic copy available at: https://ssrn.com/abstract=2575841

 missiles and the long-range missiles will kill many of State A’s
civilians before they are destroyed.
In this case, State A can prevent State B from using its long-range missiles to
kill State A’s soldiers and civilians without first destroying State B’s anti-
aircraft missiles. However, if State A first destroys State B’s anti-aircraft
missiles then fewer of State A’s civilians will be killed. Evidently, Timely
Prevention is just a special case of Triple Prevention, in which we compare the
losses inflicted by the attack on the anti-aircraft missiles with the losses
prevented by destroying the long-range missiles sooner rather than later.
Now consider the following variation on the previous cases:
Costly Triple Prevention: If State A destroys State B's anti-aircraft
missiles then some of State B's civilians will be killed as a side-effect.
However, if State A does not first destroy State B's anti-aircraft
missiles then State A’s air forces will suffer substantial losses before
destroying State B’s long-range missiles.
In this case, State A can prevent State B from using its long-range missiles to
kill State A’s soldiers and civilians without first destroying State B’s anti-
aircraft missiles. However, if State A first destroys State B’s anti-aircraft
missiles then fewer of State A’s soldiers will be killed trying to prevent State
B from using its long-range missiles. It follows that an attack on the anti-
aircraft missiles is justified only if the losses that those missiles would inflict
on attacking forces are substantially greater than the civilian losses that the
attack itself would inflict.
Arguably, cases like Costly Triple Prevention are not governed by the
proportionality principle but by the precautions principle. On this
interpretation, the true targets are the long-range missiles and attacking
forces must choose between one method of attacking those targets that
involves greater harm to civilians but less harm to attacking forces and
another method of attacking the same targets that involves greater harm to
attacking forces but less harm to civilians. As we saw in chapter seven, the
precautions principle should be interpreted to prohibit the method more
harmful to civilians unless the alternative methods would be substantially
more harmful to attacking forces.
On an alternative interpretation, there are two distinct targets in such
cases, each of which offers distinct though related military advantages. The
advantage of destroying the long-range missiles lies in the losses that this
will prevent, assuming that the attacking force will not surrender and would
otherwise pursue victory by other means. The advantage of destroying the
anti-aircraft missiles lies in the losses that this will prevent, assuming that
the attacking force will, and could proportionately, attack the long-range
missiles. On this view, Costly Triple Prevention simply combines two cases of
Simple Prevention, one nested within the other. An attack on the anti-
aircraft missiles will prevent substantially greater harm than it will inflict

Electronic copy available at: https://ssrn.com/abstract=2575841

because it will enable an attack on the long-range missiles that will in turn
prevent substantially greater harm than it will inflict.
The proper classification of such cases is conceptually interesting but
makes no practical difference. Both precautions and proportionality are best
understood in terms of the moral asymmetry between doing harm and
allowing harm and will permit or prohibit the same conduct. However, as a
matter of legal interpretation, it seems better to classify such cases under the
proportionality rule. The precautions rule governs the choice of alternative
means and methods of carrying out a single attack on a single target, or a
choice between targets. However, attacking a secondary target does not seem
like a method of attacking a primary target. Instead, we should view these as
distinct attacks on distinct targets.
As a general matter, attacks on opposing combatants, military
equipment, and military facilities will share the same causal structure as the
four preceding cases. Such attacks either prevent the object of attack from
inflicting future harm, prevent the object of attack from preventing the
attacking force from preventing some further object of attack from inflicting
future harm, prevent the object of attack from delaying the attacking force
from preventing some further object of attack from inflicting future harm, or
prevent the object of attack from inflicting harm on the attacking force or its
civilian population as it seeks to prevent some further object of attack from
inflicting future harm. Such attacks are proportionate if they (or the
sequence of attacks of which they are a necessary part) prevent (or enable the
prevention of) substantially greater harm than they inflict.
Attacks that aim to capture strategic locations introduce an additional
causal step between the initial attack and the prevention of harm. Such
attacks either indirectly prevent harm by denying such locations to the
opposing force or enable the attacking force to prevent harm in subsequent
operations. Though the causal structure of such cases is more complex, their
moral structure is the same. Such attacks are proportionate only if they
indirectly prevent (or indirectly enable the prevention of) substantially
greater harm than they (or the sequence of attacks of which they are a
necessary part) inflict.

The Value of Military Advantage

We can now see where, on my account, the value of a military advantage lies.
Let X represent the total losses that one’s own forces and civilians will suffer
in the remainder of the conflict if a military advantage — say, destroying a
munitions factory or killing an insurgent leader — is not achieved. Let Y
represent the total losses that one’s own forces and civilians will suffer in the
remainder of the conflict if that military advantage is achieved. The value of
the military advantage is the difference between X and Y, that is, the total
overall losses prevented by achieving the advantage. Let Z be the losses that

Electronic copy available at: https://ssrn.com/abstract=2575841

one would unintentionally inflict on opposing civilians in pursuit of the
advantage. On my view, an attack is proportionate just in case the difference
between X and Y is substantially greater than Z.
In more familiar terms, collateral harm to civilians is proportionate
only if it is justified by military necessity. According to one influential
formulation:

Military necessity permits a belligerent, subject to the laws of war, to

apply any amount and kind of force to compel the complete submission
of the enemy with the least possible expenditure of time, life, and
money.26

Since the complete submission of the enemy seldom depends on a single

attack, almost all attacks are justified by military necessity only to the extent
that their execution will reduce the expenditure of time, life, and money in
pursuit of military victory, relative to their non-execution. Evidently, the
conservation of time and money cannot justify collateral harm to civilians,
except insofar as shortening the conflict or conserving resources would avoid
or enable the avoidance of comparable harm to others. It follows that attacks
that collaterally harm civilians are justified by military necessity only if their
execution reduces the expenditure of life and serious injury in pursuit of
military victory, relative to their non-execution.
Put another way, the value of a military advantage is the difference in
marginal value between a broader campaign that requires achieving that
advantage and alternative campaigns that do not. Harm to civilians inflicted
in pursuit of a military advantage satisfies jus in bello proportionality only if
any alternative campaign would allow substantially greater harm to civilians
and to the attacking force. More precisely, if an attack that is part of
campaign 𝑆! inflicts degree of harm H on opposing civilians, then the attack
satisfies jus in bello proportionality only if campaigns 𝑆! through 𝑆! , which do
not include the attack, would occasion more harm to one’s own forces and
civilians than 𝑆! , and if the net difference in overall harm is substantially
greater than H. Advanced militaries already compare possible courses of
action when planning operations, typically in terms of effectiveness, risks to
attackers, and costs.27 On my view, they should also compare possible
courses of action in terms of the harms they will inflict on civilians and the
harms they will allow their adversary to inflict on attackers or on civilians.
Why should the proportionality of an attack depend on a comparison
between the campaign of which it is a part and alternative campaigns in
which it is not a part? Such a comparison is unavoidable because the harms

26US v List (American Military Tribunal, Nuremberg, 1948), 11 NMT 1230, 1253.
27See, eg, Allied Joint Publication AJP-5, Allied Joint Doctrine for Operational-Level
Planning, 26 June 2013, 3-37; US Naval War College, Workbook on Joint Operations
Planning Process, NWC 4111H, 21 January 2008.

Electronic copy available at: https://ssrn.com/abstract=2575841

that an attack will prevent are just the harms that the opposing party will
inflict if the attack is not carried out, which in turn depend on what the
attacking force will do instead of carrying out that attack: retreat, surrender,
or (more likely) design and carry out an alternative military campaign.
Again, jus in bello proportionality assumes that parties will pursue military
victory, by otherwise permissible means, at the lowest cost to their forces and
civilian population. The harm that a particular attack prevents is just the
harm that the attacking force would suffer if it pursued the next-least costly
and otherwise permissible campaign.28
Of course, it is often the case that a series of attacks overdetermines
the success of a military campaign. In such cases, each attack is causally
connected to success even though success is not counterfactually dependent
on any individual attack. In my view, an act that infringes rights can only be
justified by its results if those results are counterfactually dependent on the
act. It follows that attacks that harm civilians and make no difference to the
success or cost of a military campaign may be objectively impermissible.
However, since the results of an attack are always difficult to predict, it may
be epistemically permissible to launch a series of attacks some of which will
prove superfluous and therefore objectively impermissible. Only such a view
seems to capture the moral reality that war inevitably involves pointless
suffering that is recognizable as such only in retrospect.

In rare cases, securing a particular military advantage is strictly necessary

for military victory. In such cases, there is no campaign sufficient for
military victory that does not require securing this advantage. As we have
seen, jus in bello proportionality compares the costs of pursuing victory by
first securing a particular military advantage with the costs of pursuing
victory without first securing that advantage. It follows that an attack
violates jus in bello proportionality only if victory can be achieved without
first securing that advantage. We may say that attacks necessary to secure
advantages that are necessary for victory necessarily satisfy jus in bello
proportionality.29
Jus in bello proportionality can foreclose many ways to win, but cannot
foreclose the only way to win. Of course, ‘the rules of international law must
be followed even if it results in the loss of a battle or even a war’.30 However,

28 Note that full moral justification requires a comparison between campaigns that
each satisfies the jus ad bellum and other jus in bello norms. Certainly, an attack cannot be
morally justified on the grounds that if the attack is not carried out then the attackers will
pursue a campaign that violates other moral or legal standards.
29 Seth Lazar suggested to me that attacks that secure such necessary advantages

necessarily satisfy jus in bello proportionality because if the war aim is achieved then the
war will end, thereby avoiding all the harm that would have occurred had the war continued.
Unfortunately, wars do not always end when one party achieves its aims.
30 US v List (American Military Tribunal, Nuremberg, 1948), 11 NMT 1230, at 1272

para 128.

Electronic copy available at: https://ssrn.com/abstract=2575841

following the rule of jus in bello proportionality may result in the loss of a
battle but may not result in the loss of a war. While the jus in bello
categorically prohibits attacking civilians irrespective of anticipated military
advantage, the jus in bello only prohibits collaterally harming civilians out of
proportion to anticipated military advantage. Accordingly, civilian losses
cannot outweigh the military advantage of winning a war under the jus in
bello. However, as we shall see, civilian losses can outweigh the aims of the
war under the jus ad bellum.
Crucially, to say that attacks necessary for victory do not violate jus in
bello proportionality is not to say that such attacks are lawful let alone
morally permissible. In particular, if an attack necessary for victory would
cause more harm than victory itself would justify then the war effort as a
whole would violate jus ad bellum proportionality. Properly understood, jus
ad bellum proportionality compares the harm inflicted by an overall military
strategy, including by all of its constituent attacks, with the aims of the war
as a whole. Accordingly, attacks necessary to the success of the overall
strategy may satisfy jus in bello proportionality while the broader strategy
violates jus ad bellum proportionality. If an attack necessary for victory
would cause more harm than victory itself would justify then the problem lies
not at the jus in bello level but at the jus ad bellum level, not with the attack
as such but with the overall strategy of which it is a necessary part. In other
words, to say that an attack necessary for victory is disproportionate is just to
say that the pursuit of victory is disproportionate, and is best understood as a
jus ad bellum claim rather than as a jus in bello claim.
The legal standard of jus ad bellum proportionality is not found in any
treaty but resides instead in customary international law. Unfortunately,
state practice and opinio juris remains confused and fractured, with
substantial support for three very different approaches. On one approach,
proportionality requires that defensive force must be comparable in its scale
and effects to that of the armed attack to which it responds. On a second
approach, proportionality requires that defensive force must be no greater
than necessary to prevent or repel the armed attack to which it responds.
Finally, on a third approach, proportionality requires that the harm caused
by defensive force must be outweighed by the need to prevent or repel the
armed attack to which it responds.31 Many scholars favor the second,

31 See, eg, Dapo Akande & Thomas Liefländer, ‘Clarifying Necessity, Imminence, and
Proportionality in the Law of Self-Defense’ (2013) 107 American Journal of International
Law 563, 566; David Kretzmer, The Inherent Right to Self-Defence and Proportionality in
Jus Ad Bellum’ (2013) 24 European Journal of International Law 235, 278-79. See also Tom
Ruys, ‘Armed Attack’ and Article 51 of the UN Charter (CUP 2010) 121-22 (‘Defensive action
resulting in large numbers of civilian casualties has unvariably evoked strong negative
reactions from the international community and has frequently been condemned by the
Security Council and/or the General Assembly’). Note that the condemnations cited by Ruys
typically refer to the disproportionate use of force, invoking the jus ad bellum, not to the
disproportionate conduct of hostilities, which would invoke the jus in bello.

Electronic copy available at: https://ssrn.com/abstract=2575841

‘instrumental’ or ‘functional’ approach, often adopting a state-centric view of
the jus ad bellum while relegating humanitarian considerations to the jus in
bello.32 This view is somewhat surprising, since the object and purpose of the
modern jus ad bellum regime is ‘to save succeeding generations from the
scourge of war’.33
When states disagree over the content of a customary rule, which
interpretation should prevail, at least provisionally, until general consensus
is reached? If states disagree over the content of a prohibition then generally
the prohibition should be construed narrowly, since international law
presumptively permits (in the weak sense) what it does not clearly prohibit.34
In contrast, if states agree regarding the content of a prohibition but disagree
over the content of an exception to that prohibition then generally the
exception should be construed narrowly, since the prohibition presumptively
applies absent an agreed-upon exception. Importantly, international law
generally prohibits the use of military force, with a limited exception for self-
defense. Accordingly, if states disagree over the limits of self-defense then we
should favor the more restrictive position, namely that the resort to force,
even if commensurate with and necessary to prevent the armed attack to
which it responds, may be rendered disproportionate if the total harm it
inflicts is out of proportion to the defensive aim it seeks.

Before moving on, let me take a moment to distinguish my view from one
that McMahan effectively criticizes, namely the view that jus in bello
proportionality compares harms inflicted on civilians with harms avoided to
combatants in the very same engagement.35 On this view, just combatants
may inflict harm on civilians if such harm is a necessary and proportionate
side-effect of using defensive force against unjust combatants. This view
could at most explain the permissibility of unit self-defense, that is, of force
used to repel a direct attack on particular combatants. However, as
McMahan observes, this view would preclude all offensive operations, since if
a party refrains from offensive operations then the necessity to use defensive
force on behalf of its members often will not arise. In particular, McMahan
argues that this view would preclude humanitarian military interventions.
Since attacking forces could simply not intervene, attacking forces will inflict
losses on civilians that are not strictly necessary to protect their members. In
addition, it is not clear how this view would apply to targeted killing
operations in which the attacking force is never in danger.
My view is not subject to these objections. On my view, we hold
constant the war aim of the attacking force and ask whether an attack is a

32 See, eg, Yoram Dinstein, War, Aggression and Self-Defence (CUP 2005) 237-42.
33 United Nations Charter, preamble.
34 See SS “Lotus” (Fr v Turk), Judgment, 1927 PCIJ (ser A) No 10, at 18 (Sept. 7).
35 Jeff McMahan, ‘War Crimes and Wrongdoing in War’, in RA Duff et al (eds), The

Constitution of Criminal Law (OUP 2013).

Electronic copy available at: https://ssrn.com/abstract=2575841

necessary part of a broader campaign for achieving military victory. Attacks
are proportionate, including as part of a humanitarian intervention, only if
the immediate losses that they (or the sequence of attacks of which they are a
necessary part) inflict are substantially less than the future losses that they
will prevent over the remainder of the conflict, assuming always that the
attacking force will pursue military victory.
According to McMahan, ‘Whereas necessity requires comparisons
between an act of defense and alternative means of avoiding a threatened
harm, proportionality requires a comparison between an act of defense and
doing nothing to prevent the threatened harm’.36 On this view, it might seem
that, by tying the harm prevented by an attack to what the attacking force
would permissibly do if it does not carry out the attack, I am conflating
proportionality and necessity. Not so. It is just that, in war, there are two
threats to consider: the micro-threats posed by particular military targets
(such as infantry units or artillery positions) and the macro-threat posed by
the opposing armed force as a whole. Accordingly, if the jus ad bellum
permits a strategic response to the macro-threat posed by the opposing force
then the micro-threats posed by particular targets depend on which otherwise
permissible strategy is chosen.
For example, suppose that State A launches an unjust war against
State V. To end State A’s aggression, State V plans a land invasion of State
A, in full conformity with the jus ad bellum. Before and during the invasion,
State V attacks military targets deep inside State A that would otherwise
pose micro-threats to State V’s invasion force. Surely such attacks are
governed by the proportionality rule rather than by the necessity or
precautions rule. Yet the alternative to destroying the targets and
preventing the micro-threats they pose is hardly ‘doing nothing’. If the
targets are not destroyed then the invasion force will still invade, but will
take heavier losses in the process. Accordingly, the micro-threats posed by
the targets depend on what the invading force would permissibly do if the
targets are not destroyed.
Indeed, if the proportionality rule applied only to micro-threats that
cannot be avoided through non-forcible means, as McMahan seems to suggest,
then it would apply only to cases of unit self-defense, and perhaps to cases
like Simple Prevention, but not to offensive military campaigns that make up
much of war. Such a view seems open to many of the objections that
McMahan offered to the simple view discussed above.

36 McMahan, ‘Proportionate Defense’ 3. McMahan also writes that ‘necessity

compares the expected consequences of an act of defense with those of other means of defense,
negotiation, or retreat, while proportionality compares the consequences of an act of defense
with those of submission’. Ibid. This view seems plausible with respect to the jus ad bellum.
If force is necessary to prevent or repel an armed attack, then the only remaining decision is
to resist or submit. However, assuming that the jus ad bellum is satisfied, the remaining
decision under the jus in bello is how to resist.

Electronic copy available at: https://ssrn.com/abstract=2575841

Symmetry and Independence

On my view, jus in bello proportionality assumes that all parties to a conflict

will pursue military victory and asks whether particular attacks prevent
substantially greater harm to the soldiers and civilians of one side than they
inflict on the civilian population of another. One virtue of my approach, then,
is that it allows both participants and observers to evaluate the
proportionality of particular military operations without reference to the
justice or legality of the broader conflict. A party that violates the jus ad
bellum may still conform to jus in bello proportionality. Conversely, a party
that conforms to the jus ad bellum may systematically violate jus in bello
proportionality.
The law assumes that warring parties will seek to pursue military
victory at the least possible cost to their own soldiers and civilians. Jus ad
bellum proportionality constrains each party’s pursuit of military victory by
comparing the value of their war aims with the total cost of their pursuit. In
contrast, jus in bello proportionality constrains a party’s attempts to reduce
its own losses by comparing the future harm that an attack prevents with the
immediate harm that it inflicts. Accordingly, a military strategy may satisfy
jus ad bellum proportionality while many of its constituent attacks violate jus
in bello proportionality. Alternatively, a military strategy may violate jus ad
bellum proportionality while many of its constituent attacks satisfy jus in
bello proportionality.
Certainly, the law requires both the independence of jus in bello
proportionality from jus ad bellum considerations and the symmetrical
application of jus in bello proportionality to all sides of a conflict. Presumably,
a legal rule that flatly prohibits harming civilians in pursuit of an unjust
cause will be ignored both by combatants who believe that their cause is just
and by combatants who feel compelled to fight for a cause that they recognize
is unjust.
Not everyone will see the symmetrical application of my approach as a
virtue. Recall that Hurka and McMahan argue that a military advantage
that does not contribute to the achievement of a just cause has no moral
value that could justify the moral disvalue of harm to civilians. They
conclude that any harm inflicted on civilians by unjust combatants in pursuit
of an unjust cause is disproportionate from a moral point of view.
Hurka and McMahan are clearly right that attacks by unjust
combatants are seldom fully morally justified, since they generally kill and
injure just combatants, make the achievement of unjust causes more likely,
and are in one sense unnecessary if the opposing party would agree to a
ceasefire and ultimately to a just peace. However, attacks by unjust
combatants may be less wrongful to the extent that they prevent harm to
others. Since almost all civilians retain their moral right not to be harmed,
there are almost always strong moral reasons to prevent civilians from being

Electronic copy available at: https://ssrn.com/abstract=2575841

harmed. Similarly, since even unjust combatants do not deserve death or
dismemberment, there is almost always some moral reason to prevent them
from being harmed. An attack is less wrongful to the extent that it is
supported by such reasons and more wrongful to the extent that it is not.
Since proportionality is in part a function of such reasons, it is almost always
morally worse for unjust combatants to harm civilians in violation of jus in
bello proportionality than for them to harm civilians without violating jus in
bello proportionality. It follows that even combatants who knowingly fight
without a just cause (perhaps under duress) have significant moral reasons to
comply with jus in bello proportionality. In doing so they will seldom act
permissibly, but they will almost always act less wrongfully than they would
otherwise.
There is an obvious objection to this argument. Suppose that State A
is an unjust aggressor, that State B is fighting solely in national self-defense,
and that if State A stops fighting then State B will stop fighting as well. On
these facts, State A could prevent all future harm to its own soldiers and
civilians without harming any of State B’s civilians simply by ending its
aggression. It follows that State A is not even partially morally justified in
harming any of State B’s civilians. This argument succeeds subject to an
important qualification. State A’s political and military leaders are indeed
responsible for all the harm inflicted by their armed forces, including in
operations that do not violate the law of armed conflict. If they are
prosecuted for the crime of aggression then their sentences should reflect the
total number of civilians and combatants harmed in pursuit of their unjust
cause. These leaders may not claim jus in bello proportionality in defense of
their conduct or in mitigation of their moral fault.
However, it is hardly ever in the power of operational and tactical
commanders, let alone ordinary soldiers, to end an unjust war. They cannot
do so on their own and are unlikely to succeed in mounting an effective joint
or collective effort. As a result, unjust combatants often must choose between
harming just combatants and foreign civilians or allowing just combatants to
harm their fellow soldiers and their own civilians. In many such cases,
individual unjust combatants are partially morally justified in defending
their fellow soldiers and their own civilians. Often, unjust combatants should
refuse to fight on balance, that is, often their reasons to refuse are even
stronger than their reasons to fight. However, their reasons to fight are
powerful — even if defeated by even more powerful opposing reasons — and
it is in this sense that they are partially justified in fighting.37
It is true that, on my view, jus in bello proportionality does not prohibit
unjust combatants from collaterally harming just combatants or civilians in

37 Cf Vera Bergelson, ‘Rights, Wrongs, and Comparative Justifications’ (2007) 28

Cardozo Law Review 2481, 2493 (‘A partially justified act is, therefore, a wrongful act that,
due to certain mitigating circumstances, is less wrongful than that required by the charged
offense’).

Electronic copy available at: https://ssrn.com/abstract=2575841

order to prevent harm to unjust combatants (themselves included). Yet many
just combatants do not pose unjust threats and therefore are not liable to
defensive harm. Moreover, many unjust combatants pose unjust threats and
therefore are liable to defensive harm. It follows that it is often objectively
impermissible for unjust combatants to collaterally harm either just
combatants or civilians in defense of unjust combatants. This inescapable
fact seems to favor an asymmetric standard.
At the same time, as we saw in chapter three, many just combatants
pose unjust threats to civilians who retain their rights not to be harmed.
These just combatants may be liable to defensive harm—perhaps even if the
unjust threats they pose to civilians are themselves justified under the
proportionality rule. Accordingly, it may be permissible for unjust
combatants to collaterally harm such just combatants in defense of such
civilians—perhaps even if doing so would prevent the prevention of
substantially greater harm to others. Moreover, it may be permissible for
unjust combatants to collaterally harm such just combatants, as well as
civilians, in defense of unjust combatants, if only so that these unjust
combatants may defend innocent civilians from unjust threats in the future.
Of course, not all just combatants will pose unjust threats to civilians,
and not all unjust combatants will defend civilians from unjust threats. It
follows that a gap will remain between symmetrical law and asymmetrical
morality. This gap may be justifiable, or at least tolerable, for pragmatic and
epistemic reasons. After all, most combatants believe that they are just
combatants and that their opponents are unjust combatants. Accordingly, in
practice, most combatants would simply ignore an asymmetric standard. As
a final plea in mitigation, we should recall that the law of armed conflict is
prohibitive rather than permissive. Accordingly, the law of armed conflict
prohibits acts that violate jus in bello proportionality, but does not permit,
authorize, or condone acts that conform to jus in bello proportionality. Acts
not prohibited by the law of armed conflict may be prohibited by other legal
or moral norms.
As we saw in chapter two, the moral function of the law of armed
conflict is to help all combatants, just and unjust alike, conform to the moral
reasons that apply to them. For unjust combatants, full conformity is
generally impossible and partial conformity is the most that they can achieve
so long as they continue to fight at all. However, given the moral stakes of
armed conflict, even partial conformity is most welcome. If unjust
combatants stop harming civilians except as a side-effect of preventing
substantially greater harm to civilians or combatants then armed conflict will
be much less unjust than if they are not so constrained.

Expected Losses and Anticipated Advantages

Electronic copy available at: https://ssrn.com/abstract=2575841

As we have seen, the proportionality principle requires attacking forces to
predict the harm that an attack may inflict on civilians, typically in the form
of collateral damage estimates (CDEs). On my view, attacking forces must
also predict the harm that their attacks will prevent over the remainder of
the conflict. Such predictions may seem impractical but in fact are
ubiquitous in warfare and reflect a basic skill of responsible command. Most
notably, commanders regularly make such predictions when deciding to carry
out attacks that place their own forces in danger. Commanders seek to
achieve military victory at the least possible cost to their own forces and
civilians. Commanders will therefore place their own forces at risk only if
they believe that doing so will prevent greater harm to their own forces and
civilians over the remainder of the conflict. My view simply requires
commanders to bring together these two predictions in order to determine the
epistemic permissibility of an attack.
In general, an action is epistemically permissible only if the actor has
decisive reason to believe (a) that the action will infringe no rights, (b) that a
justifying circumstance exists, or (c) that a justifying result will occur.
Conversely, an action is epistemically impermissible if the actor has decisive
reason to believe (a) that the action will infringe rights, (b) that no justifying
circumstance exists, and (c) that no justifying result will occur. It follows
that an attack is epistemically impermissible if the attacker has decisive
reason to believe that the attack will harm civilians (an infringement of
unforfeited rights) unless the agent has decisive reason to believe that the
attack will prevent substantially greater harm to friendly forces or to friendly
civilians (a justifying result). In other words,

Probably Proportionate: An attack may be carried out only if, based on

the information reasonably accessible to the attacker, the attack will
probably prevent substantially more harm to soldiers and civilians
over the remainder of the conflict than it will inflict on civilians.

In this formulation, the degree of evidential probability reflects the strength

of the attacker’s epistemic reasons for belief. An attack is probably
proportionate if the attacker’s reasons to believe that it will prove objectively
proportionate are stronger than the attacker’s reasons to believe that it will
prove objectively disproportionate.
In many cases, there will be an amount of harm that an attack will
probably inflict — for example, using a missile that will probably kill
everyone in its blast radius — and an amount of harm that the attack will
probably prevent — for example, by providing close air support to infantry
units under fire.38 However, often there will be no single number that an

38 Cf Commentary on the HPCR Manual, (‘‘Expected’ collateral damage and

‘anticipated’ military advantage, for these purposes, mean that that outcome is probable, i.e.
more likely than not’).

Electronic copy available at: https://ssrn.com/abstract=2575841

attack will probably kill or that an attack will probably save. For example,
suppose that there is a 40% chance that an attack will kill 10, a 30% chance
that it will kill 20, and a 30% chance that it will kill 30. Similarly, suppose
there is a 40% chance that the attack will save 20, a 30% chance that it will
save 40, and a 30% chance that it will save 60. Assume that these
probabilities are independent of each other, such that the number who will be
killed is unrelated to the number that will be saved. There is no exact harm
that the attack will probably prevent and no exact harm that the attack will
probably inflict. Instead, there are nine possible outcomes, none of which is
very likely to occur. We can illustrate this by adapting the Risk
Assessment/Tolerance Matrix used in NATO operational planning doctrine:39

.4 harm 10 .3 harm 20 .3 harm 30

.4 save 20 .16 (20:10) .12 (20:20) .12 (20:30)
.3 save 40 .12 (40:10) .09 (40:20) .09 (40:30)
.3 save 60 .12 (60:10) .09 (60:20) .09 (60:30)

Nevertheless, we can still determine whether or not the attack will probably
result in a proportionate outcome. For example, there is only a 33% chance
that the attack will prevent at least three times more harm than it will inflict.
These possible outcomes are indicated in bold. It follows that if harming is at
least three times worse than allowing harm then the attack will probably not
prove proportionate and is therefore epistemically impermissible.
Alternatively, there is an additional 34% chance that the attack will prevent
twice the harm that it will inflict. These possible outcomes are indicated in
italics. It follows that if harming is less than twice as bad as allowing harm
then there is a 67% chance that the attack will prove proportionate and is
therefore epistemically permissible. If the possible harm inflicted and the
possible harm prevented are not independent in this way then the attacking
force would have to determine the likelihood of each possible outcome and
then determine whether the attack will probably result in one of the
proportionate outcomes.
Any serious account of jus in bello proportionality will require
attackers to make similar probabilistic judgments with respect to both
civilian losses and military advantages, however the latter are understood.
Nevertheless, such probabilistic judgments may seem too complicated for
attacking forces to perform in many situations. Often the probabilities
themselves are uncertain or the possible outcomes too numerous to carefully
examine. In such cases, attacking forces may be unable to determine the
epistemic permissibility — let alone the objective permissibility — of their
actions. Attacking forces will have no choice but to base their decisions on
presumptions, heuristics, and other mental shortcuts.

39 Allied Joint Doctrine for Operational-Level Planning, 3-39.

Electronic copy available at: https://ssrn.com/abstract=2575841

 Since attacking forces tend to underestimate the harm that they will
inflict and overestimate the harm that they will prevent, they should adopt a
decision procedure that skews in the opposite direction. I propose the
following:

Max-Min: If an attacker cannot determine whether or not an attack

will probably prevent substantially more harm than it will inflict then
the attack may be carried out only if the minimum harm that the
attack might prevent is substantially greater than the maximum harm
that the attack might inflict.

To exclude very unlikely outcomes, the term ‘might’ stands in for a

probability threshold, say 20%, above which possible harm becomes operative.
For example, if there is a 20%-or-greater chance that an attack will harm as
many as 30 civilians and a 20%-or-greater chance that the attack will save as
few as 20 soldiers or civilians, then the attack should not be carried out. If
attackers cannot determine that an attack will probably prove proportionate
then they should assume the worst — within limits — and act accordingly.
One might think that an attack is permissible if the expectable harm to
civilians — that is, the average of the possible harms to civilians discounted
by their respective probabilities — is substantially less than the expectable
military advantage — that is, the average of the possible military advantages
discounted by their respective probabilities. In chapter six, we described this
concept as expectable proportionality. The attack described above would
expectably-harm around six and expectably-save around 12. The attack is
therefore expectably proportionate if doing harm is at most twice as bad as
allowing harm but expectably disproportionate otherwise. In this case, my
approach and the expectabilist approach yield the same result. However, in
other cases the two approaches diverge.
For example, suppose that a targeted killing operation directed at a
particular low-level insurgent will almost certainly harm several civilians
and prevent little or no harm to soldiers because such fighters are seldom
effective and so easily replaced. However, there is a small chance that this
low-level insurgent will one day develop into a senior leader who will make a
necessary contribution to the harming of many soldiers. Even if the harm
that the attack will expectably prevent is substantially greater than the
harm that the attack will expectably inflict, the attack is epistemically
impermissible because it will almost certainly inflict far greater harm than it
will prevent. If the attack ultimately harms civilians and saves no one then
the attacker could not possibly defend the attack to those civilians.
Here, as elsewhere, reasonable belief — that is, belief supported by
decisive epistemic reasons — sets a minimum threshold of epistemic
permissibility. However, we can endorse more restrictive standards on
expectabilist grounds. For example, even an attack that will probably

Electronic copy available at: https://ssrn.com/abstract=2575841

prevent substantially greater harm than it will inflict is expectably
impermissible if it will inflict greater expectable harm than it will prevent.
For example, if an attack carries a low probability (say 10%) of inflicting very
great harm (say 1000 civilian deaths) and no probability of preventing very
great harm then it may be expectably impermissible even if it will probably
prevent greater harm (say 30 combatant deaths) than it will inflict (say 10
civilian deaths). Thus, an attack should be prohibited if it is either probably
disproportionate or expectably disproportionate.
Finally, the reasonable belief threshold is no more absolute than other
deontological constraints. For example, even an attack that will probably not
prevent substantially greater harm than it will inflict is permissible if it will
prevent far greater expected harm than it will inflict. For example, if an
attack carries a low probability (say 10%) of preventing very great harm (say
1000 civilian deaths) and inflicting much less harm (say 10 civilian deaths)
then it may be epistemically permissible even if it will probably prevent only
slightly more harm (say 20 civilian deaths) than it will inflict (say 10 civilian
deaths). Thus, an attack is permissible only if it is either probably
proportionate or expectably overriding.

Moral Uncertainty

On my view, proportionality incorporates the moral asymmetry between

doing harm and allowing harm. Expected harm to civilians is excessive in
relation to anticipated military advantage when the former exceeds what the
latter would justify. The value of a military advantage lies in the harm to
combatants and civilians that its achievement will prevent over the
remainder of the conflict. Since doing harm is substantially worse than
allowing harm, it follows that expected harm to civilians exceeds what
anticipated military advantage would justify when the latter will not prevent
substantially greater harm to others.
As we saw at the beginning of this chapter, proportionality judgments
are somewhat imprecise. For example, we rarely judge that an attack would
be proportionate if it will likely harm N civilians but would be
disproportionate if it will likely harm N+1 civilians. Most scholars suppose
that proportionality judgments are imprecise because they compare
incommensurable values, namely civilian losses and military advantage. Of
course, I have argued that the value of a military advantage lies in the losses
prevented by its achievement. Thus, on my account, proportionality
judgments compare fully commensurable values, namely harms to some and
harms to others.
On my view, proportionality judgments are imprecise because the
moral asymmetry between doing and allowing harm is imprecise. When we
say that killing is substantially worse than letting die, we do not mean that
killing is precisely X or Y times worse than letting die, where X and Y are

Electronic copy available at: https://ssrn.com/abstract=2575841

exact numbers like 3.14. Rather, we mean that killing is at least X times
worse and at most Y times worse than letting die, where X and Y are
substantial figures. Put another way, we mean that collaterally killing one
innocent person would be a disproportionate side effect of preventing fewer
than X innocent people from being killed but a proportionate side effect of
preventing more than Y innocent people from being killed. At the same time,
the permissibility of collaterally killing one innocent person as a side-effect of
saving between X and Y innocent people may be impossible to determine.
Importantly, on my view, the imprecision of our proportionality
judgments is significant but hardly debilitating. Certainly, it would be
disproportionate to collaterally harm one innocent person as a side-effect of
preventing comparable harm to only one other innocent person. This much
seems to follow from any nontrivial asymmetry between doing harm and
allowing harm. Yet even requiring a 1:2 ratio of harm inflicted to harm
prevented would transform military practice. After all, the imprecision of the
proportionality rule is thought by many to render it either fairly permissive
or very permissive. In contrast, on my account, the imprecision of the
proportionality rule renders it either fairly restrictive or very restrictive.
With respect to the upper bound (Y), responses to the famous Trolley
Problem suggest that it would be proportionate to collaterally harm one
innocent person as a side-effect of preventing comparable harm to five other
innocent people.40 It is true that the Trolley Problem involves redirecting an
existing threat, while collateral harming in war involves creating new threats
as a side-effect of eliminating existing threats. Arguably, it is harder to
justify creating new threats than to justify redirecting existing threats. At
the same time, it is possible that most people would think it permissible to
redirect a threat away from two or three innocents toward one innocent.
Accordingly, most people might think it permissible to create a new threat
that would collaterally harm one innocent to prevent comparable harm to five
innocent persons from an existing threat. Empirical studies of lay and expert
judgments on these matters would be most illuminating.
Since proportionality judgments are necessarily somewhat imprecise,
we should ask whether the law of armed conflict should prohibit attacks that
are neither clearly proportionate nor clearly disproportionate. First, we
should distinguish how attacking forces should act and how reviewing
tribunals should judge. Obviously, reviewing tribunals should only punish
individuals if the legal element of excessiveness is established beyond
reasonable doubt. As we shall see in chapter ten, this is the truth contained
in the view that criminal liability should only attach to attacks that are
clearly excessive.41

40 See, eg, Helen Frowe, ‘Claims Rights, Duties and Lesser Evil Obligations’ (2015) 89

Proceedings of the Aristotelian Society, Supplementary Volume, 267, 277; McMahan,

‘Proportionality and Necessity in Jus in Bello’.
41 See Rome Statute, art 8(2)(b)(iv).

Electronic copy available at: https://ssrn.com/abstract=2575841

 However, if attacking forces cannot determine whether the military
advantage anticipated from an attack would justify the expected civilian
losses then they should refrain from attack. As the ICRC comments,

the disproportion between losses and damages caused and the military
advantages anticipated raises a delicate problem; in some situations
there will be no room for doubt, while in other situations there may be
reason for hesitation. In such situations [of doubt] the interests of the
civilian population should prevail.42

Harming civilians is not wrong because it is disproportionate. On the

contrary, harming civilians is wrong unless it is proportionate. Attacking
forces should presume that collaterally harming civilians is wrong unless
they acquire decisive reason to believe that they are justified in doing so.
Accordingly, combatants should refrain from attacks that are not clearly
proportionate.

Moral Standards and Decision Procedures

In war, as elsewhere, often the surest way to miss a target is to aim at it

directly, ignoring both adverse conditions as well as our own imperfections.
Instead, we may be more likely to hit our target by aiming a bit above or
below, to the left or to the right. Similarly, we may be more likely to satisfy a
moral standard not by trying to apply it directly but by applying a decision
procedure that anticipates and corrects for potentially distorting factors. The
standard of jus in bello proportionality that I propose is a complex one.
Attackers may better conform to that standard indirectly, by following a
decision procedure that relies as much on intuition as on calculation.
In the regulation of armed conflict, the distinction between moral
standards and decision procedures corresponds to the distinction between
legal rules and rules of engagement (ROE). ROE do not directly track
applicable legal rules in all of their complexity but instead provide clear and
straightforward instructions to combatants of varying ages and education
levels that, if followed, will more often succeed in eliciting lawful conduct.
To help combatants conform to jus in bello proportionality, I propose
two ROE. First, consider

Us = Them: Do not carry out an attack that will endanger civilians

unless you would accept the same risk to your own unit to achieve the
same goal.

42 ICRC, Protocol I Commentary para 1979.

Electronic copy available at: https://ssrn.com/abstract=2575841

 Ask yourself: Will the attack harm civilians? If so, would you carry
out the attack even if your unit would suffer the same amount of
harm? If you would not, then DO NOT carry out the attack.

To illustrate, Us = Them would direct an officer not to order a missile strike

against an insurgent that would collaterally kill three civilians unless the
officer would be willing to lose three soldiers in order to kill the insurgent.
The officer might consider whether he would order a ground assault that
would claim the lives of three soldiers, or whether he would order the strike
even if three soldiers, held hostage by the insurgent, would be collaterally
killed in the strike. In general, officers will endanger their own troops now
only to substantially reduce danger to their own troops or civilians later. It
follows that most officers who directly follow Us = Them will indirectly
conform to jus in bello proportionality.
Next, consider the following ROE:

Ours = Theirs: Do not carry out an attack that will endanger foreign
civilians unless you would accept the same risk to your own civilians.

Ask yourself: Will the attack harm civilians? If so, would you carry
out the attack even if they were your own civilians? If you would not,
then DO NOT carry out the attack.

To illustrate using the same scenario, Ours = Theirs would direct a

commander not to order a missile strike against an insurgent that will
collaterally kill three foreign civilians unless the commander would order the
strike even if it would instead collaterally kill three of his own civilians. In
general, officers will endanger their own civilians now to substantially reduce
danger to their own troops or civilians later. It follows that most officers who
directly follow Ours = Theirs will indirectly conform to jus in bello
proportionality.
As we saw in the introduction, soldiers have special duties to protect
their fellow soldiers and fellow citizens even at grave risk and at great cost to
themselves. However, these special duties do not give soldiers special rights
to harm foreign civilians as a means or as a side-effect of protecting their
fellow soldiers or fellow citizens. Soldiers must bear the costs of their own
special duties and may not impose costs on foreign civilians that the latter
have no duty to accept. It follows that soldiers may not impose costs on
foreign civilians that they may not impose on their fellow soldiers or fellow
citizens. Accordingly, my proposed ROE require that attacking forces take
losses to foreign civilians just as seriously as they take losses to their fellow
soldiers and fellow citizens. Since attacking forces too often take the former
much less seriously than the latter, these ROE serve a valuable corrective
function in moral deliberation.

Electronic copy available at: https://ssrn.com/abstract=2575841

 Importantly, military officers routinely decide how many soldiers
under their command they are willing to risk or lose in order to obtain a
given military advantage. Us = Them simply prohibits these officers from
imposing greater risks or losses on civilians in order to obtain the same
military advantage. With respect to Ours = Theirs, Ian Henderson reports
that

I have used this simple concept [that collateral harm to foreign

civilians is proportionate only if comparable harm to fellow civilians
would be proportionate] in lectures, exercises and operations and can
attest to its usefulness in helping military members ‘grasp’ the concept
of applying proportionality rather than merely thinking about it in
abstract terms.43

Hopefully, these proposed ROE will provide military officers with practical
decision procedures that are simpler and more direct than the proportionality
principle itself.
Of course, these ROE are only decision procedures and not moral or
legal standards. For one thing, much like the Golden Rule, these ROE
depend for their application on the subjective values of the applicant. The
less value an officer places on the lives of his own forces, the less value these
ROE require him to place on the lives of civilians. These ROE only tell us to
place equal value on the lives of our own forces and the lives of civilians.
They do not tell us how much absolute value to place on the lives of anyone.
At the same time, since most attackers — consciously or subconsciously —
place as much or more value on their troops than on foreign civilians, most
attackers will act less wrongfully if they follow these ROE than if they allow
their implicit biases to affect their targeting decisions.

Superiors and Subordinates

Until now we have discussed the duties of attacking forces as collective

agents without attending to the moral division of labor within such collectives.
However, the moral and legal duties of superiors and subordinates differ in
potentially significant ways. One obvious difference is that operational
commanders must ensure the proportionality of entire military campaigns,
while tactical commanders must ensure the proportionality of particular
attacks. More interesting differences arise between those who order attacks
and those who carry them out.
For example, Protocol I requires ‘those who plan or decide upon an
attack . . . [to] refrain from deciding to launch any attack’ that would inflict
disproportionate harm on civilians.44 This norm clearly applies only to those

43 Henderson 229.
44 Protocol I art 57(2)(a)(iii).

Electronic copy available at: https://ssrn.com/abstract=2575841

who design and order military operations. However, Protocol I also directs
that ‘an attack shall be cancelled or suspended if it becomes apparent’ that
the attack will inflict excessive harm on civilians.45 The use of the passive
voice suggests that this norm applies to all soldiers at all levels, including
those who only carry out attacks. According to the ICRC, ‘[t]he rule set out
here . . . applies not only to those planning or deciding upon attacks, but also
and primarily, to those executing them’.46 Indeed, if it is apparent to a
subordinate that an attack would be disproportionate then the attack is
manifestly unlawful and any order by a superior to carry out the attack must
be refused.47
True, subordinates will not always know the value of the military
advantage that an attack would yield and therefore cannot always know
when new information regarding possible harm to civilians reveals the attack
to be disproportionate. In many cases, subordinates will better conform to
their moral obligations by deferring to their superiors, whose access to
information, analysis, and legal counsel better positions them to make
reliable proportionality judgments. If the attack is in fact disproportionate,
then responsibility lies with the superiors and not with the subordinates.
Importantly, however, there are at least three situations in which
subordinates should not defer to their superiors.48 First, subordinates may
learn that the possible harm to civilians is much greater than their superiors
anticipated. Presumably, a judgment based on false factual assumptions
warrants little deference. Accordingly, such subordinates should not carry
out the attack unless they relay the new information to their superiors and
then receive orders to proceed with the attack. Second, subordinates may
have reason to doubt that their superiors are applying the proportionality
standard reasonably and in good faith. Accordingly, such subordinates
should not carry out attacks unless the information provided by their
superiors convinces the subordinates themselves of their lawfulness. Finally,
subordinates may conclude, based on the information available to them, that
an attack appears so grossly disproportionate as to be manifestly unlawful.49
In such extreme cases, it is so unlikely that additional evidence, available
only to their superiors, would support the contrary conclusion that
subordinates may not defer but must insist on additional information before
carrying out an attack.
For example, suppose that a drone pilot is ordered to launch a missile
at an unnamed, low-level insurgent away from an active battlefield. Before
launching the missile, the pilot sees a dozen civilians pass into the expected

45 Protocol I art 57(2)(b); ICRC, Customary IHL Study 60.

46 ICRC, Protocol I Commentary para 2220. See also Quéguiner 803.
47 See Rome Statute of the International Criminal Court, art 33(1)(c) (superior orders

are not a defense to war crimes if the defendant did not believe that the orders were lawful or
if the orders were manifestly unlawful).
48 Cf Raz, The Morality of Freedom 42-46.
49 Ibid.

Electronic copy available at: https://ssrn.com/abstract=2575841

blast radius of the missile. Presumably, the potential harm to civilians is
much greater than the pilot’s superiors anticipated. Moreover, the attack
now appears so grossly disproportionate that it is highly unlikely that the
pilot’s superiors have specific information regarding the targeted individual
that would support a judgment of proportionality. It is very unlikely that the
targeted individual will otherwise kill substantially more than a dozen
soldiers or civilians, as would be required to render the attack proportionate.
In this case, the pilot legally must and morally should suspend the attack.50

Targeted Killings Revisited

Thus far, I have illustrated my account of jus in bello proportionality mostly

using skeletal cases involving international armed conflict. While the US–
Iraq war shows that state-to-state conflict is hardly a thing of the past, it is
worth pausing to apply my account to non-international armed conflicts
between states and non-state actors.
For over a decade, the United States has pursued a strategy of
targeted killing of individual adversaries using unmanned aerial vehicles
(UAVs) or ‘drones’. One of the supposed virtues of drones is that they can
hover over a target for an extended period and strike when few or no civilians
are in harm’s way. Indeed, the ratio of civilians killed to combatants killed
by drone strikes has fallen from almost 1:1 under the Bush Administration to
under 1:4 under the Obama Administration and continues to fall.51 However,
the Obama Administration also expanded the drone campaign from targeting
mostly high-level terrorists and insurgents to targeting primarily mid-level
and low-level Taliban fighters. As a result, the number of drone strikes
increased dramatically in 2010, and the annual number of civilian deaths
remained very high for several years. These changes in the design and
execution of the drone program make it both more difficult and more
important to evaluate its morality and legality.

50 Cf Commentary on the HPCR Manual 94 (‘Anyone with the ability and authority to
suspend, abort or cancel an attack, must do so once he reaches the conclusion that the
expected collateral damage would be excessive in relation to the anticipated military
advantage. For instance, a pilot who has the target in view and unexpectedly observes
civilians in the target area — who were not supposed to be there, based on the information
provided to him during the briefing preceding the attack — must assess the collateral
damage expected to befall them and cancel the attack if he concludes that the principle of
proportionality will be violated’); Quéguiner 805 (‘[I]f, before launching a first salvo against a
bridge, a tank driver notices that a crowd of fleeing civilians have taken refuge under the
targeted bridge, the driver cannot assume that the planners have correctly considered the
principle of proportionality and continue his mission in wilful blindness and impunity. He
must, at the very least, suspend his attack in order to allow the civilians to evacuate, or to
request that his orders be confirmed in the light of these new circumstances’).
51 See Bureau of Investigative Journalism, ‘Covert Drone War’,

http://www.thebureauinvestigates.com/category/projects/drones/.

Electronic copy available at: https://ssrn.com/abstract=2575841

 To evaluate the proportionality of a particular drone strike, we must
compare the losses that the strike will inflict on civilians to the losses that
the strike will prevent from befalling soldiers or civilians over the remainder
of the conflict. It therefore seems that drone strikes against low-level
insurgents and terrorists that unintentionally kill civilians will seldom prove
proportionate. Many low-level insurgents will never kill anyone, and many
are so easily replaceable that killing them will not prevent substantially
greater future harm. Such insurgents may be liable to be killed even if they
will be replaced if killed. However, attacks that inflict losses on civilians can
be proportionate only if the attacks also prevent substantially greater harm
to others. It follows that if killing a low-level insurgent will not prevent
substantially greater harm to others then it would be disproportionate to
harm any civilians in the process.52
In contrast, a high-level insurgent commander might make a necessary
contribution to the deaths of many soldiers or civilians over the remainder of
a conflict. High-level insurgents may be difficult to quickly replace, or may
be replaced by others with less skill or charisma. In principle, it could be
proportionate to collaterally kill some civilians as a side-effect of killing such
a high-level insurgent.53 Importantly, even if a particular drone strike
satisfies jus in bello proportionality, the overall strategy of using drones to
achieve broader war aims may violate the jus ad bellum.

Conclusion

An attack that inflicts incidental harm on civilians is proportionate only if it

prevents substantially greater harm to the attacking force or to its civilians
over the remainder of the conflict. This account of jus in bello proportionality
does not compare incommensurable values, only immediate losses to civilians
and future losses to civilians and to attacking forces. In addition, this
account applies symmetrically to all parties to an armed conflict,
independently of the jus ad bellum morality and legality of their use of
military force. Attacks that are disproportionate under this account are
morally impermissible when carried out by just combatants, and

52 For a similar conclusion reached on strategic rather than moral or legal grounds,
see David Petraeus, U.S. Dep’t of Def., The U.S. Army and Marine Corps Counterinsurgency
Field Manual, para 7-32 (U.S. Army Field Manual No. 3-24, 2006) (‘In COIN environments,
the number of civilian lives lost and property destroyed needs to be measured against how
much harm the targeted insurgent could do if allowed to escape. If the target in question is
relatively inconsequential, then proportionality requires combatants to forego severe action,
or seek non-combative means of engagement’).
53 Cf Michael N Schmitt, ‘Unmanned Combat Aircraft Systems and International

Humanitarian Law: Simplifying the Oft Benighted Debate’ (2012) 30 BU Int’l LJ 595, 616
(‘Multiple civilian casualties may not be excessive when attacking a senior leader of the
enemy forces, but even a single civilian casualty may be excessive if the enemy soldiers killed
are of little importance or pose no threat’).

Electronic copy available at: https://ssrn.com/abstract=2575841

disproportionate attacks carried out by unjust combatants are morally worse
than proportionate attacks carried out by unjust combatants. It follows that
both just and unjust combatants have decisive moral reasons to avoid attacks
that are disproportionate under this account, and the law would guide
soldiers well by prohibiting such attacks.

Electronic copy available at: https://ssrn.com/abstract=2575841

y prohibiting such attacks.

Electronic copy available at: https://ssrn.com/abstract=2575841