The notion that some means of waging war are mala in se is a confronting one. Surely, any weapon can be used for good or ill? Philosophers often try to justify the category of mala in se by suggesting that some weapons are inherently incapable of being used in accordance with the just war principles of distinction and proportionality. This line of argument faces two obvious objections. First, claims about the limits of particular weapons typically fail to consider the different contexts in which they might be used. For example, anti-personnel mines can be used as defensive measures for fixed installations in marked locations that are fenced off from civilian intrusion. Second, deriving the category of mala in se from the principles of distinction and proportionality makes it redundant. The argument that some weapons cause disproportionate suffering is more persuasive, but it falters if one conducts an honest appraisal of the nature and extent of the suffering caused by weapons that are not held to be mala in se.

Thus, if we want to hold that some weapons are evil in themselves, we have two choices. One, we can ground claims in the realm of meanings rather than of consequences. When we choose our weapons we reveal something about ourselves: if we fail to acknowledge the humanity of our enemy we may jeopardize our own. Alternatively, we can understand the list of weapons we have chosen to prohibit as a historical achievement. The very existence of this list represents a choice by the international community to marginally reduce the horror of war by turning away from certain means of fighting it.

I offer these brief—and inevitably controversial—reflections on the category of mala in se here because I believe many of the criticisms Jenkins and Purves make of my article “Robots and Respect” ultimately stem from a deeper disagreement about its usefulness.Footnote ¹ First, however, let me pause here to acknowledge how much agreement there is between me and Jenkins and Purves. In particular, we agree that there is a large class of machines (mostly hypothetical, for the moment) that could be plausibly categorized as autonomous weapon systems (AWS) and that would not rule out an interpersonal relationship between the person who launches them and the people they kill. Unlike Jenkins and Purves, however, I continue to worry that there might one day be a class of still more sophisticated machines—artificial agents, with “wills” of their own—that would vitiate this relationship, not because they would introduce “metaphysical indeterminacy” about who they will kill, but because the agency of the robot would intervene so as to make it implausible to describe the person who launched the robot as having killed anyone at all. This is the “responsibility gap,” which, as Jenkins and Purves note, is not a new issue.Footnote ² In any case, the primary focus of my article was the ethics of the former category of AWS, not least because real-world AWS will fall within this category for the foreseeable future.

While such weapons would maintain an interpersonal relationship between the killer and the killed, I suggested that this would be a relation of disrespect rather than respect and cited the depth of public feeling about autonomous weapons in support of this claim. Jenkins and Purves dismiss this move summarily as a concession to public ignorance and irrational fears. However, insofar as whether a particular way of treating people is respectful or disrespectful depends on what such treatment means, references to widely shared understandings are entirely appropriate. Killing people with robots would be disrespectful in (roughly) the same way that—but to a much higher degree than—spitting or urinating on people would be or, as I suggested in my article, mutilating their corpses would be.

Jenkins and Purves argue, following Ron Arkin, that a concern for respect for noncombatants may actually require us to embrace AWS should robots eventually prove able to cause fewer noncombatant casualties than human warfighters. However, as Jenkins and Purves anticipate, I want to suggest that if a weapon is malum in se, the fact that employing it would save the lives of third parties is irrelevant.Footnote ³ Perhaps they mean that if the use of AWS were motivated by a concern for the rights of noncombatants, it would not be disrespectful of the humanity of combatants. Yet this does not follow. The meaning of person A's treatment of person B is not determined by A's intentions; nor does the fact that A's disrespectful treatment of B benefits others mean that the action is no longer disrespectful. Interestingly, we can see here another way in which the attempt to account for the category of mala in se in terms of distinction and (wide) proportionality is problematic: some means of warfighting are evil because of the way they treat their intended targets and not because of—indeed, regardless of—their implications for noncombatants.

The argument outlined above—and more fully in the original article—is weaker than many critics of AWS, including myself, would prefer. Claims about the meaning of technologies are notoriously contested. The implications of existing weapons for the bodies of those they target and for the relationship between killer and killed are also often horrific. As people become more familiar with robots in civilian life, the intuition that they would be a dehumanizing means of killing may weaken, especially if the “humanitarian” case for AWS becomes compelling. Let me therefore conclude by saying something briefly in support of the second way of understanding the category of mala in se mentioned above—as a contingent historical achievement—and its relevance to AWS.

As I have argued elsewhere, AWS have the potential to radically change the nature of warfare, as well as to significantly lower the threshold of conflict.Footnote ⁴ I believe that these jus ad bellum concerns provide an indirect justification for declaring AWS to be mala in se. Unless the international community acts swiftly to prohibit the development and deployment of AWS, they will be developed.Footnote ⁵ Once developed, they will be deployed. Once deployed, they will quickly come to dominate the way wars are fought, with the result that the future of nations will be determined by the decisions of machines. Agreeing now to categorize such machines as mala in se may be the only way to resist this dynamic and its dangerous consequences. Even if Jenkins and Purves are correct, and the argument from respect is weaker than I had hoped, its potential contribution to this urgent task justifies its exploration.