When is it ok to purposely kill civilians?

On August 6th, 1945, the United States dropped a nuclear bomb on Hiroshima. Tens of thousands of men, women, and children were instantly vaporized. Others were charred into ghoulish forms. Radiation slaughtered many more through the slow rotting and peeling of their flesh, and the introduction of leukemia into their veins.¹ Truman argued it saved millions more from the brutalization of war. History has debated whether it was either a permissible albeit regrettable act of destruction or terrorism carrying the dove’s flag. 

First, we must define the terms. Civilians are those who are not in the armed forces, though they can contribute to harm (e.g., corporate funders of an unjust war).² To target civilians is to harm (often kill) them intentionally. (The section entitled ‘the uncertainty principle’ addresses their identification as civilians.) Terrorism is impermissibly doing so. Moreover, because your membership in some group has no bearing on the permissibility of your actions, impermissible actions are as such regardless of whether their perpetrators are states.³ No uniform can justify terrorism. 

The tenet of my argument is that no one moral theory can capture the world’s nuance. The consequences of actions are never certain. Sometimes harm is justified. Civilian contributions to harm come in varying sizes and degrees of directness. Civilians of different moral status are often harmed in one stroke. The victims of war and oppression do not deserve oversimplification. 

As for my argument, I first cover what we are permitted to do under uncertainty. I then respond to consequentialist, self-defense-based, and doctrine-of-double-effect-based arguments. I argue that we are only permitted to target civilians who are a substantial contributor to unjust harm. And, if other kinds of civilians are harmed/killed as a foreseen yet unintended side effect, the benefit must vastly outweigh their harm. 

The uncertainty principle

Gathering facts is hard and time-consuming enough. Using facts to predict the future is harder still, for uncountable factors influence long-term outcomes. Hence, most actions are taken under uncertainty. Such actions fall into two categories. The first are those that must be taken immediately. As such, these actions are permissible only if the available relevant facts suggest they are. For instance, someone walking towards troops while clutching a gun-shaped object may be liable to harm. The second category comprises actions that need not be done immediately. The same principle applies, but the minimum amount of relevant facts is proportional to the moral risk. Though being short of gathering all the relevant facts is permissible if the moral risk is low (e.g., when considering which enemy compound to target), the higher the moral risk (e.g., when choosing whether to target a compound that could be an enemy holdout or a wedding), the more relevant facts are needed. Crucially, such a ‘moral risk’ principle prevents groups from invoking plausible deniability after doing wrong when they could have done more research. 

“It [saved] hundreds of thousands of lives” - Harry S. Truman⁴

Consequentialists argue it is permissible to target x civilians to save y > x civilians. That is false, because doing harm is much worse than allowing harm. Take the following two scenarios, for example: (1) Alice allows Eve to die; (2) Bob kills Eve. In (1), Alice is allowing harm, which is defined in a slightly complicated manner—most of the things Alice could have done would not have involved saving Eve’s life. In (2), meanwhile, Bob is harming; likewise, most of the things Bob could have done would not have involved killing Eve.⁵ Notice (2) is z times worse than (1), as it is z times worse to kill someone than it is not to save them. It is therefore wrong to target x civilians to save the lives of y > x civilians; rather, doing so is only permissible when y > xz. Here, any z may seem arbitrary; however, for practical purposes, it is prohibitively large. Indeed, as many would agree, you cannot permissibly torture a few innocent civilians to save a city’s worth. As that yields an astronomically high z, which has not yet been matched in history, speculation about z’s exact value is useless for the essay question. 

A slightly contorted variation of the argument yields that a group's failure to save the lives of its members is worse than its targeting of non-members. (Maybe that is the least the group could do given it limits its members' freedoms and resources, has promised to help them, or is its members’ main constitutive attachment.) Consider the following counter-example, however: the only way you can save the life of your starving child is to kill your neighbor, steal his food, and feed your child. Though you might consider your failure to feed your child worse than your failure to feed, say, your neighbor, you may not permissibly kill your neighbor. As such, it is still impermissible to kill x civilians to save y > x members of your group. Plausibly, however, z may be somewhat smaller. 

The fact that actions have uncertain consequences weakens both of the previous arguments, too. It is near-certain that by targeting x civilians, roughly x civilians will die. But it is far less certain that you will save y > x civilians and/or group members because: (1) the prospect is often long-term, which means it is influenced by a coterie of confounding factors; (2) even if it is short-term, it often relies on the behavior of others (e.g. killing some civilians to prevent the enemy from killing more), which is inherently more uncertain. The fact that the probability of y civilians and/or group members being saved is far less than one means y has a vastly smaller expected value.

“We are acting in self‑defense.” — IDF Chief of Staff Eyal Zamir⁶

The strongest non-consequentialist justification for targeting civilians is that we are permitted to do so in self/other defense. 

Before proceeding, we must briefly discuss the issue of culpability. Introducing it into the fray makes the activity of determining liability so complicated that I will stipulate its non-existence for the sake of brevity. That said, the standard argument against the existence of culpability denies the existence of free will. Scientists are in near consensus that we are only made of particles, which behave according to the laws of the universe, or may, in some circumstances, which physicists can specify, behave randomly. Either way, we control none of our behavior, emotions, and thoughts, and therefore lack culpability.⁷ Naysayers respond that being unable to act otherwise does not preclude culpability; a man whose actions are controlled by a demon such that he must harm, yet, by his deliberation, would have harmed anyway, is culpable for his decision.⁸ Reality, however, is more closely aligned to a scenario in which the demon controls every part and moment of the man’s thinking; intuitively, he is no longer culpable.

Two principles unequivocally do apply to self/other-defense, however. First, we are only permitted to inflict proportionate defensive harm.⁹ Regarding the essay question, such harms are probably death, grievous injury to body and/or property, and/or severe oppression. Second, defensive harm is only permissible if it is necessary.¹⁰ If agents could be culpable, that may not be true; we are permitted to kill culpable killers even if we could stop them by merely harming them—culpability does not exist, however.

Bearing these principles in mind, three categories exist in which civilians are contributing to harm, the latter of which is defined using the aforementioned ‘most of the things you could have done’ account.¹¹

Substantial contributor

Examples include armament workers, corporate funders, key scientists, spies, and prominent propagandists—those on whom the harm’s continuation depends.¹² Interestingly, culpability bears no relevance here. Take a sleepwalker attempting to kill, for instance. Although he is non-culpable, he is liable. 

Partial contributor

Paradigm examples include the voter and the taxpayer. If they were liable, it would result in the unintuitive conclusion that we are permitted to kill entire voting and taxpaying populations. Troublingly, however, consider a scenario in which many sleepwalking henchmen each inject a small amount of poison into a man; if they all do so, the man will die. It seems permissible for him to kill (at least) one contributor to escape; the two scenarios do not seem to be different, for a voter and a taxpayer plausibly contribute to at least one death. The difference, however, lies in the actors’ directness of contribution. The henchmens’ contributions directly cause the man’s death. Yet only after many causal steps will the actions of the voter and/or taxpayer contribute to harm. Crucially, a causal step is defined as causing a different type of action. For instance, in a scenario where the president issues a hit order which passes through a general before reaching soldiers, both the president and the general are issuing orders, while the soldiers are doing the killing—the causal step lies between the general and the soldiers. Paying taxes, to take one example, must go through a tax agency, which solicits funds for the government, which then gives those funds to armament workers, who then load weapons onto trucks, whose drivers deliver those weapons to soldiers. Taxpaying initiates a domino effect that comprises too many different action types for the taxpayer to be liable. Although the degree of removal beyond which someone is non-liable is unclear (and is therefore outside the scope of my argument), armament workers and weapons transporters fall on one side, and taxpayers and voters on the other. Hence, even without comprehensive theorizing, we know it is impermissible to kill partial contributors whose actions are far removed from the harm. 

Justified contributor

As is being discussed, harm is justified if it is proportionate, necessary, and inflicted against liable persons. As such, contributors to these justified harms are not themselves liable. For instance, Russia, which is waging an unjust war, cannot invoke self-defense to kill Ukrainian civilians who fight back.¹³ 

The last two categories concern civilians who are not contributing to harm.

Bystander

It is unequivocally impermissible to target bystanders.¹⁴ Indeed, it is abhorrent to defend yourself from a knife-wielding assailant by dragging over a passing child to use as a shield. Bystanders who support unjust harm are nonetheless illegitimate targets because of their lack of culpability.

Omitters

Some civilians are justified in omission (see definition of justification above). They are clearly non-liable. However, some civilians may not be causally implicated in unjust harm, but may nonetheless fail to mitigate/prevent it; such civilians include those who live in a democratic society, yet, at little cost to themselves, fail to engage in protest that would be likely to stop their government’s waging of an unjust war.¹⁵ Since these civilians are non-culpable, however, they are non-liable. 

Mixed categories

Often, many civilians are killed at once, some of whom may be liable to defensive harm, and others not. In such situations, the Doctrine of Double Effect (DDE) suggests we are permitted to kill the non-liable civilians as a foreseen but unintended side effect of killing the liable.¹⁶ The first condition for these actions to be permissible is that they are necessary for the benefit. The second is that they result in much greater benefit. Yes, it is only permissible to target x civilians to save y > xz lives; but in DDE situations, z is much smaller, as the difference between intentionally doing harm and intentionally allowing it is much greater than that between accidentally doing harm and accidentally allowing it. Indeed, to save many more lives, it is impermissible to level a city, yet permissible to bomb armament factories with nearby residential areas. And, as the uncertainty principle suggests, we should err on the side of caution if y > xz lives may not actually be saved. 

In summary, civilians can only be the main target if the uncertainty principle suggests: (1) they are substantial contributors to unjust harm; (2) the harmful side effect (if it exists) is vastly outweighed by the benefit.

Endnotes

1.  “Hiroshima and Nagasaki Bombings.” n.d. ICAN. https://www.icanw.org/hiroshima_and_nagasaki_bombings.

2.  “Customary IHL - Rule 5. Definition of Civilians.” n.d. https://ihl-databases.icrc.org/en/customary-ihl/v1/rule5.

3.   McMahan, Jeff. Killing in war. Oxford, NY: Clarendon Press ; Oxford Univ. Press, 2009. 

4.  “The President’s Farewell Address to the American People | Harry S. Truman.” n.d. https://www.trumanlibrary.gov/library/public-papers/378/presidents-farewell-address-american-people

5.  Bennett, Jonathan, 1967, “Acting and Refraining,” Analysis, 28: 30–31.

6.  Fayyad, Abdallah. 2025. “What Does Israel’s ‘Right to Self-defense’ Actually Mean?” Vox, March 18, 2025. https://www.vox.com/world-politics/403719/israel-right-to-self-defense-gaza-palestine-international-law

7.  Sapolsky, Robert M., Determined: A Science of Life Without Free Will. Penguin Press, 2023.

8.  Fischer, John Martin, and Mark Ravizza, 1998. Responsibility and Control: A Theory of Moral Responsibility, Cambridge: Cambridge University Press.

9.  McMahan, Jeff. Killing in war. Oxford, NY: Clarendon Press ; Oxford Univ. Press, 2009. 

10.  Ibid.

11.  Here, we are not discussing counterfactual actions in the sense of free will, but rather plausible counterfactuals; even if you could not have chosen otherwise, causally close counterfactual worlds exist in which you take different actions.

12.  McMahan, Jeff. Killing in war. Oxford, NY: Clarendon Press ; Oxford Univ. Press, 2009. 

13.  Ibid.

14.  Ibid.

15.  Ibid.

16.  Aquinas, Thomas (13th c). Summa Theologica II-II, Q. 64, art. 7, “Of Killing”, in On Law, Morality, and Politics, William P. Baumgarth and Richard J. Regan, S.J. (eds.), Indianapolis/Cambridge: Hackett Publishing Co., 1988: 226–7.

References

“Hiroshima and Nagasaki Bombings.” n.d. ICAN. https://www.icanw.org/hiroshima_and_nagasaki_bombings.

“Customary IHL - Rule 5. Definition of Civilians.” n.d. https://ihl-databases.icrc.org/en/customary-ihl/v1/rule5. 

McMahan, Jeff. Killing in war. Oxford, NY: Clarendon Press ; Oxford Univ. Press, 2009.  

“The President’s Farewell Address to the American People | Harry S. Truman.” n.d. https://www.trumanlibrary.gov/library/public-papers/378/presidents-farewell-address-american-people 

Bennett, Jonathan, 1967, “Acting and Refraining,” Analysis, 28: 30–31.

Fayyad, Abdallah. 2025. “What Does Israel’s ‘Right to Self-defense’ Actually Mean?” Vox, March 18, 2025. https://www.vox.com/world-politics/403719/israel-right-to-self-defense-gaza-palestine-international-law

Sapolsky, Robert M., Determined: A Science of Life Without Free Will. Penguin Press, 2023.

Fischer, John Martin, and Mark Ravizza, 1998. Responsibility and Control: A Theory of Moral Responsibility, Cambridge: Cambridge University Press.

Aquinas, Thomas (13th c). Summa Theologica II-II, Q. 64, art. 7, “Of Killing”, in On Law, Morality, and Politics, William P. Baumgarth and Richard J. Regan, S.J. (eds.), Indianapolis/Cambridge: Hackett Publishing Co., 1988: 226–7.