Technology does not inherently make war more ethical. Nevertheless, this is the claim often put forward by proponents of machine learning, artificial intelligence (AI), and autonomous systems—including political, military, and corporate actors. Throughout modern history, ethics has been mobilized in relation to technological changes to argue that war can be made more ethical, more moral, and more humane through technological refinements. In the interwar years, airpower was described by U.S. Air Force officer Harold Arnold as “the most humane of all weapons.”Footnote 1 First Winston Churchill in 1919, and then the United States during Vietnam, justified the use of gas in “moral” and “humanitarian” terms.Footnote 2 Tactical nuclear weapons have been claimed to “make U.S. deterrence both more ethical and more effective” due to greater precision and credibility of threat.Footnote 3 Proponents have argued that the MQ9 Reaper drone is “the most potentially ethical use of airpower yet devised.”Footnote 4 Much of the discourse around drones has characterized them as “a step forward in humanitarian weapons technology,”Footnote 5 able to provide for “a more moral campaign.”Footnote 6 Technological advances have thus long been seen as solutions to ethical problems in war. Embedded in that thinking is a teleological belief in linear progress, that we are marching toward the most ethical form of warfare through technological prowess. AI, machine learning, and autonomous systems are only the latest iteration of technologically enabled “ethical war,” seen as finally able to fulfill the dream of “ethical weapons”Footnote 7—whether this be through greater precision, the presumption that ethics and international law can be effectively coded within machines, or the belief that machines can reign in the worst instincts or mistakes of humanity.Footnote 8 Military applications of AI can, allegedly, “improve selectivity, accuracy, accountability, and traceability” and therefore “make armed conflicts less unethical and inhumane.”Footnote 9 There is also a perceived need for Western states to pursue these ethical technologies in order to maintain strategic advantage in a world in which they see themselves as the ethical leaders competing with their unethical and undemocratic adversaries. The presumption that technology will make war more ethical and that Western states are the ethical actors able to achieve such a vision plays a significant role in thinking about the ethics of war and military technology. To contest this assumption it is important to foreground the politics and power involved in constructing ethics discourse in Western states.
To do so it is important to understand more precisely how new technologies are shaping how we think about ethics and war through better understanding the ideas underlying these technologies. The starting point for this article is the following question: What impact do the intellectual inheritances of machine learning and AI research have on how Western states approach the ethics of war? Research on machine learning and AI is underpinned by a Cartesian and Newtonian philosophical foundation.Footnote 10 Ideas such as rationality, certainty, control, prediction, order, and the existence of objective, neutral, and universal knowledge are central to machine learning research and development. These ideas do not simply disappear when these technologies enter the military domain. Rather, when the intellectual foundations and ideologies associated with machine learning research reach the military domain they intersect with existing ways of thinking about ethics and the experience of war. Many aspects of war ethics scholarship and military thought rely on similar intellectual foundations: scientific and technological ways of thinking have long shaped military thought, including with regard to certainty, rationality, control, prediction, order, and chaos.Footnote 11 Various elements of just war traditions and international legal obligations underpin (at least theoretically) defense and state approaches to war ethics. Meanwhile, a fixation on precision and technological advancements has simultaneously reified dreams of ethical and humane war, while shifting the focus of ethics from the nature of any particular war as a whole, to the limited realm of the targeting process.Footnote 12 Across all of these different traditions, ethics has never been neutral and it has always done political work that obscures both the politics and the experience of war.
The aim of this article is neither to argue that there is no such thing as ethics nor that ethics is always epiphenomenal to factors such as capital and identity. Rather, it is to demonstrate that the way ethics discourse is being advanced in relation to military technologies in Western states is not representative of any kind of neutral or universal form of ethics, but is politically contextual and influenced by factors including capital and identity. As Maja Zehfuss argues, “ethics and politics are indeed always already intertwined.”Footnote 13 Ethics discourse and practice cannot take place outside the realm of politics because the context in which ethics is discussed and deployed is already political.Footnote 14 Drawing on the work of Derrida, Zehfuss critiques the treatment of ethics as a domain separate from politics. Rather than seeing ethics as somehow “pure” or “separate,” both they and I start with the assumption that ethics and politics are always intermingled.Footnote 15 It is therefore important to interrogate the politics and interests that sit behind how the language of ethics is being used by defense forces and the defense industry in relation to new and emerging technologies to understand what role ethics is playing in these particular contexts.
The intersection of the intellectual foundations of machine learning with military thought results in a strengthening of certain pre-existing tendencies in thinking about ethics and war. Here I shall focus on two that I consider to be dominant: ethics as code and ethics as identity. By “ethics as code,” I mean the drive to simplify and quantify ethics, most recently into machine compatible formats. By “ethics as identity,” I mean the reduction of complex ethical debates to a simple belief that “we” are the ethical actors and the “other” is not. Neither of these is new, but they are being reinforced in the context of machine learning, AI, and autonomous systems. To demonstrate these trends, I will draw on examples from across Western states, focused on the United States but drawing on insights from other states including the United Kingdom, Australia, and Israel.Footnote 16 The United States remains the global hegemon and is the driver of many of the relevant technological developments and identity narratives, with the U.K. and Australia playing important roles through their alliances and partnerships with the United States. Israel is a key site of technological development not only for itself but also global defense manufacturers, having a significant influence not only on technological development but also the ideas and identity narratives that come with those technologies.Footnote 17 It is important to focus on how the discourse of ethics is used, what discourses appear in relation to ethics, and what material interests are attached to them. We know that ethics has often been invoked in ways that exacerbate rather than constrain violence, deployed as a justification for waging war in different contexts based on ideas of what ethical war looks like and who can perform it.Footnote 18 Both manifestations of ethics analyzed here, ethics as code and ethics as identity, push the experience of war to one side, either through ideas of ethical and humane war via improved technological capabilities or through the belief that we are ethical actors and therefore the wars we pursue and how we pursue them will be ethically justified.
In this article, the experience of war haunts in its relative absence. It is not a central concern for those purveying ideas of technologically enabled ethical warfare. Ethics as code and ethics as identity are about how and when to conduct war, rather than how to be subjected to war. These forms of ethics center the purveyors of war and not its subjects. War becomes decontextualized and depoliticized. In this context, rather than debate the ethics and politics of when war should be waged, the focus in much of war ethics discourse is on how to codify and how to be, by virtue of nationality and technological capability, ethical. Despite the prominence of ethics in Western military thought, what happens instead of ethical war is simply more war, with the very idea of the possibility of an ethical war used to justify waging war in more and more contexts rather than constraining war.Footnote 19 It is for this reason that attention needs to shift away from questions of how to make new technological systems ethical or what it would take for them to comply with international law and toward more anti-war positions.Footnote 20 This involves a two-part focus on the experience of war. First, to foreground the experience of those subjected to war, to reject narratives of the possibility of ethical or humane war. Second, to foreground the experience of the makers of war—by which I mean the expertise and capital deployed to create conditions for, and advocate in favor of, war.
The Intellectual Inheritances of Machine Learning
Technology comes with certain intellectual inheritances. It is never neutral or universal, but always political and particular. When it comes to AI and machine learning, Abeba Birhane has argued that the intellectual inheritances influencing the development and dissemination of these technologies are Cartesian and Newtonian. By this, she means Descartes’s focus on uncovering “the permanent structures beneath the changeable and fluctuating phenomena of nature in which he could build the edifice of unshakeable foundations of knowledge,” thereby creating the rational individual.Footnote 21 Birhane argues that while there are few who subscribe to a wholehearted Cartesian view of the world, elements of it linger and facilitate a view of the world that privileges “complete understanding, control, order, manipulation, and prediction.”Footnote 22 From the Newtonian tradition she draws out the drive to “impose order and to arrive at universal and objective knowledge.”Footnote 23 However, these ideas about the physical world as “containing discrete, independent, and isolated atoms” do not translate well “to the messy, interactive, fluid, and ambiguous world of complex living systems that are inherently context bound, socially embedded, and in continual flux.”Footnote 24 James Scott makes a similar argument regarding what he terms “high modernism,” long before the technological systems under consideration here existed. In the same way that Birhane argues that Cartesian and Newtonian logics simplify reality, Scott argues that states set about simplifying their social realities to make them legible, including through measures such as the census.Footnote 25 AI is not only or even mostly driven by states trying to simplify their realities, but this desire to operate as though there is a universal set of mappable truths is one of the key worldviews shaping its emergence, use, and regulation—albeit one more driven by Silicon Valley than the state. Having realized some of the uses for big data the state has become greedy for it, and that data has become a cornerstone of how states see their worlds.Footnote 26
The impact of these intellectual inheritances and associated values is not simply theoretical. They get embedded into the research done and technologies produced. For instance, analysis of existing machine learning research has found that the kinds of values that receive attention in academic research are those such as performance and efficiency rather than user rights, autonomy, or justice.Footnote 27 Similarly in defense science research there is a focus on simple legal compliance rather than broader moral or ethical principles and, even more so, the willingness of defense personnel to use or trust new technologies (rather than a consideration of whether these technologies should be used).Footnote 28
Current machine learning systems epitomize the type of technological outcomes that come from relying on these intellectual traditions. For starters, they rely on the presumption of universality and neutrality.Footnote 29 Purveyors of AI have succeeded in many realms of social life at marketing their products in such terms, obfuscating the politics and power at the heart of their creations. Politics and power are, however, infused throughout every possible layer of AI, machine learning, and autonomous systems. Power and politics sit at the heart of decisions made in classifying the data used to train algorithms.Footnote 30 They are at the heart of the production, the necessary labor, and in the extraction of both data and raw materials.Footnote 31 They are the key to the control over computing and cloud infrastructure, and the exertion of power and control over the environmental resources of energy generation and water supply required to sustain them. Further, there are certain ideological predilections shaping the development of AI and machine learning. Timnit Gebru and Émile Torres have labelled these the TESCREAL (transhumanism, extropianism, singularity, cosmism, rationalism, effective altruism, and longtermism) bundle that includes, among other things, a eugenicist underpinning.Footnote 32 While not all developers subscribe to all or even any of these ideologies, they are prominent among the powerful in Silicon Valley—including Elon Musk, Sam Altman, and Peter Thiel—and influential in American political thought.Footnote 33 These ideologies are constantly changing and evolving, spawning new variations with different focuses. While analyzing these ideologies is beyond the scope of this article, what is salient in this context is the recognition that there are various ideological stakes in the development of AI, machine learning, and autonomous systems, and that these are also tied to economic and political interests. It is important to take note of the power, politics, intellectual inheritances, and ideologies inherent in technological development before these technologies fully enter the military domain, as they are not simply expungable once they reach the battlefield. They will shape military thought, practices, and technology adoption. I explore here the implications of these foundations and how they interact with military imaginaries relating to the role of technology in war and ethics.
Rationality, Certainty, and Universality in War, Technology, and Ethics
While the above section details some of the underlying intellectual traditions of machine learning, this section brings in some of the underlying intellectual traditions of warfare, focusing on the intersection of war, technology, and ethics. These perspectives are infused with similar traditions to those Birhane identifies in machine learning research, drawing on a history of thought shaped by positivism and rationality. In combination, they push Western states toward ethics as code and ethics as identity perspectives and practices that occlude the politics of ethics, shift attention to the technical means of waging war rather than the experience of those subjected to it, and generate technoscientific imaginaries of ethical and humane war.
Technology plays a central role in ethics of war debates, grounded in a presumption that technological advances can make war more ethical or more humane. How technology is understood in relation to war is shaped by broader traditions of ethics in war, particularly just war thinking. While just war theory is a diverse field, much of the discourse, including in military thought, is grounded in the same focus on rationality, objectivity, and universality as machine learning research. Just war theory dominates how war is thought about—by scholars, practitioners, politicians, defense forces, and international law.Footnote 34 Zehfuss points out that much of just war thinking treats ethics as distinct from politics, believing that “interest, power, and indeed practicalities should not interfere with determining the ethical course of action.”Footnote 35 Ethics is seen as existing in its own realm, “distinct and pure,” even though it is impossible to achieve such a cleavage of the political and the ethical.Footnote 36 Cian O’Driscoll has outlined the way in which much of just war thinking has, through excising the experience of war, created an illusion that it is objective, impartial, and universal.Footnote 37 This dominant type of just war thinking is “the product of a set of practices (for example, detachment, abstraction) based on freighted Cartesian assumptions that reduce everything to bland hypotheticals and bloodless generalities.”Footnote 38 Elke Schwarz argues the same, criticizing the dominance of applied ethics in just war theory debates as well as international politics more broadly. Applied ethics is an approach grounded in rationality, which sees “ethics” as a fixed and acquirable outcome able to be achieved through instrumental methods such as regulatory frameworks, laws, and ethical codes.Footnote 39 It is an approach easily adaptable to an algorithmic context in which ethics are treated as something to be uncovered via scientific methods and mathematical logic.
The interconnection of technology and ethics in war is not new, but has taken on particular forms in the context of algorithmic technologies, resulting in “a discourse of scientism, objectivity, and techno-rationality that purports to make warfare inherently more ethical by virtue of its utilization of technologically advanced modes of killing.”Footnote 40 The 1991 Gulf War showcased precision warfare through technological advances in bombing.Footnote 41 As precision bombing developed over time, the fact that it was technologically possible to hit specific targets meant that in the West there was seen to be a moral and political imperative to reduce or avoid collateral damage.Footnote 42 Algorithmic targeting, AI, and autonomous systems have been presumed by proponents of their adoption to continue these trends. Autonomous weapons have been described as “ethical weapons,” with hopes of simply programming the laws of war into the technologies.Footnote 43 The picture of them painted by proponents is as “inherently rational, predictable, and even ethical.”Footnote 44 Focusing on lethal autonomous weapons specifically, Schwarz argues that debates on the morality of LAWS deploy technical modes of reasoning, “reflecting a desire to justify possible harm through abstract, algorithmic reasoning.”Footnote 45 Ethics is treated like a science with easily categorizable variables able to be folded neatly into mathematical equations or algorithmics.Footnote 46 Here we can see both the logics of war ethics and machine learning coming together, underwritten by a belief in rationality, abstraction, universality, and objectivity.
However, there is considerable critique of the argument that technology can make war inherently more ethical or humane, particularly within critical perspectives in international relations such as critical military studies, critical security studies, and the intersection with science and technology studies. The argument in this article is grounded in the work of Maja Zehfuss on the concept of ethical war. Zehfuss outlines the rise of this concept beginning with the rise of humanitarian interventions conducted by Western states in the 1990s.Footnote 47 She analyzes how the language of ethics and humanitarianism were mobilized to justify war, and how that was then adapted in a post–9/11 environment to a point where “the boundary between apparently altruistic and apparently self-interested operations is no longer recognized in the production of Western war as good.”Footnote 48 A key point she emphasizes is that the dominance of the language of ethics and enthusiasm for ethics by policymakers, military thinkers, and academics alike does not mean that ethics has actually served to constrain war in practice.Footnote 49 Zehfuss argues that this fixation on ethics has not lessened war, but has rather “served to legitimize war and even to enhance its violence.”Footnote 50 It is precisely because ethics has so often been treated as something separate from politics—above and beyond the political—that it has been able to serve a political purpose in justifying, legitimizing, and changing when, why, and how Western states wage war.Footnote 51 When war is “made through claims about ethics,” it is essential to unpack those ethical claims, their underlying intellectual inheritances, their politics, and the political economy shaping their articulation.Footnote 52
Technology has played a significant role in the construction of ethical war, including precision bombing, drones, AI, and autonomous systems. In regard to algorithmic technologies, John Emery argues that “algorithms function to discursively replace due care with a techno-ethics of war that purports a fantasy of control over the inherent uncertainties of conflict.”Footnote 53 Emery demonstrates this process through examples, one of which is the algorithmic software package colloquially referred to as Bugsplat used during the early weeks of the Iraq War in 2003.Footnote 54 Bugsplat was the iteration of collateral damage estimation tools developed by the United States for use in Iraq, building on earlier tools used from the Gulf War onwards. It was a simplification of earlier tools that enabled quicker operations during counterinsurgency compared to earlier strategic bombing campaigns, meaning it simply provided the estimated blast radius of any proposed strike.Footnote 55 Bugsplat set “an arbitrary ceiling of 30 civilian casualties,” over which the proposed strike would be further reviewed; however, the strike could still go ahead if casualty estimates were over the ceiling and the ceiling was the same for all proposed strikes regardless of their designated level of military necessity.Footnote 56 Western militaries have argued that the use of such technologies is more precise and reduces civilian casualties, thereby ensuring that “warfare would be more ethical.”Footnote 57 However, in this context “numerical ‘objectivity’ via techno-innovation is held up as an ethical end itself”—regardless of intent or outcome.Footnote 58 It is for these reasons that Eyal Weizman argues the role of civilian casualty estimates has more to do with constructing a mirage of moral and legal justification.Footnote 59 As Neil Renic and Elke Schwarz point out, the argument that such technologies can make war more ethical only holds if war actually is a rational and certain endeavor, and if ethics can be easily fitted to match this simplistic environment.Footnote 60
Both within casualty estimation tools and algorithmic war more broadly we can see a belief that ethics and the laws of war can be neutrally, objectively, and rationally applied within and through algorithmic technologies.Footnote 61 These beliefs mirror the Cartesian and Newtonian undercurrents of machine learning research, including a fixation on certainty and rationality. The development of algorithmic technologies has fueled the military dream to defeat the fog of war and acquire a “transparent battlefield”Footnote 62 allowing for “perfect situational awareness.”Footnote 63 At the core of this desire is a belief in the possibility of certainty and omniscience through technological means.Footnote 64 Transparency, in this context, is a “means of control.”Footnote 65 Similar to Scott’s Seeing like a State, we see in the context of war that transparency operates “as part of an ideology that strives to render the world visible, measurable, indexed and recorded.”Footnote 66 Practices and technologies of warfare have developed in accordance with the perceived need for greater certainty and transparency. Technologies allow for more to be seen and drive an ever-expanding appetite to see yet more—to acquire and process more data and expand surveillance capacities. Meanwhile, bureaucratic and operational processes within militaries, such as those of the United States, have developed to match the desire for transparency.Footnote 67 The desire for transparency and for certainty reaches into the future:
The desire to predict and create the future as a threat constantly demands flows of more information, more visibility and more knowledge. Moreover, it constantly demands the means (such as weaponry, education, technology) to maintain this demand.Footnote 68
Algorithmic technologies and the intellectual underpinnings of machine learning reinforce this future-oriented drive for certainty and transparency. They aim to create “a world which is transparent, open-access, logical, [and] measurable.”Footnote 69 They rely on techniques such as classification, object recognition, and pattern identification, with AI reliant on the “classification and codification of life into computable data to identify objects, and patterns between objects.”Footnote 70 The belief that ethics can be neutrally, universally, objectively, and rationally applied via technological means reinforces the treatment of ethics as code and ethics as identity, as shall be explored further below.
Ethics as Identity
The ethics-as-identity lens reduces complex ethical debates to a belief that “we” are the ethical actors while the “other” is not. It is, in many ways, part of the construction of the distinction between friend and foe for Western states. This distinction has changed in nature over time but in recent years tends to appear within a framing of democratic versus authoritarian states. When it comes to introducing algorithmic technologies into warfighting, it has been noticeable that “ethical” has become a tagline added to technologies.Footnote 71 Western militaries focus on how to adopt “ethical AI” as in Australia’s A Method for Ethical AI in Defence. Footnote 72 It is therefore important to consider what political work ideas of ethics and ethical technologies are doing.Footnote 73 As Jeremy Moses and Geoffrey Ford argue, it is precisely the necessity of the distinction between friend and foe to war that “negates the possibility of neutral and objective application of the ethics and laws of war through computational techniques.”Footnote 74 In this section I will explore the role ethics plays in identity-construction for Western states, the role technology plays in this process, and the political economy of “virtue” that shapes how algorithmic technologies are understood and their material proliferation.
Western states have identity-based narratives about ethics, constructing narratives of themselves as ethical actors opposing unethical adversaries. James Eastwood, for example, uses the Israeli Defense Forces (IDF) as an example of how “ethics can function as a weapon of war.”Footnote 75 As ethics has “become an increasingly important terrain on which war and military activity is understood, legitimated, and contested,” this turn has, Eastwood argues, favored military violence over contestation of that violence—helping to bolster the myth of Israel as “the most moral army in the world.”Footnote 76 This is not a dissimilar argument to Samuel Moyn’s argument that opposition to war has shifted from opposing war itself to a fixation on making it more humane.Footnote 77 Where Eastwood focuses on Israel and Moyn the United States, both speak to the importance that perceptions of ethical and humane war play in identity. In the case of Israel, there is widespread belief in the myth of humane war held by many among the public, military personnel, philosophers, educators, and even activists.Footnote 78 In the United States, belief in the possibility of humane war is one strongly held even by some anti-war activists and NGOs, with a shift in focus from pacificism to humane war solidified in particular in the post-Vietnam context.Footnote 79 This belief manifests itself in the campaign opposing lethal autonomous weapons systems. Drawing heavily on the work of Moyn, Moses outlines how ideas of humane war manifest in humanitarian opposition to LAWS, focusing in particular on the argument that LAWS should be opposed on the grounds of human dignity. Focusing on human dignity in this way, Moses argues, functions to reify the idea of humane and “dignified war.”Footnote 80 In the context of LAWS, even those opposing their development are therefore potentially “helping to foster a legal or normative order that serves the perpetuation of war.”Footnote 81 Zehfuss also makes this point more broadly, arguing that the centrality of ethics to Western ideas and practices of war means that “war is made through ethics” given that “the commitment to ethics enables war and indeed enhances its violence.”Footnote 82 Belief in the possibility of ethical or humane war has served to justify war waged in more places. She describes ethics as “branding”Footnote 83—it is both the performance but also the belief in ethical, good, just, dignified, or humane war. The performance and the belief both shape Western practices of warfare, the West’s approach to new technologies of war, and the political economy and politics of expertise sitting behind weapons technologies and strategic narratives.
In recent decades, technology has played a significant role in this construction of a narrative of ethicality. It is not just precise technological capabilities that have been constructed as supposedly being ethical but also Western warfare writ large.Footnote 84 When it comes to contemplating military uses of AI and autonomous systems, “machine ethics [are] justified in terms of the unjust enemy … ensuring that the military advantage realised through lethal applications of AI are executed on the basis of a moral self-image.”Footnote 85 These kinds of identity narratives are easy to find in defense documents from across Western states and in quotes from military figures, as will be demonstrated below. They each reveal identity as a key component of what will make the use of these technologies ethical.Footnote 86 The U.S. Navy, for example, has argued that ethics is a source of “competitive advantage,”Footnote 87 allowing it to foster democratic norms and influence the international standards shaping AI development and adoption.Footnote 88 Both technological superiority and “normative leadership” are seen as essential.Footnote 89 This view is contrasted against perceived fears of an “ethics gap” that could hinder technological development, as compared to less ethical and therefore unhindered adversaries.Footnote 90 In its own eyes, the United States has constructed a narrative of itself as inherently more ethical than its adversaries. One Air Force general, for instance, argued that the “Judeo-Christian” foundation of the United States meant it had better ethical codes in place for AI than its adversaries.Footnote 91 The former U.S. Deputy Secretary of Defense Kathleen Hicks has argued that “our use of AI reflects our ethics and our democratic values.”Footnote 92 In response to earlier resistance at tech companies to being overly involved in military contracts (as with the Project Maven backlash for instance, which saw employee resistance to Google’s involvement with military applications of AI), the chairman of the Joint Chiefs of Staff, General Joseph Dunford, said, “I’m not sure that people at Google will enjoy a world order that is informed by the norms and standards of Russia or China … We are the good guys.”Footnote 93 Through arguments such as these and many others, the United States is crafted as the necessary and ethical global hegemon that needs to maintain technological advantage over its unethical adversaries.Footnote 94
U.K. and Australian defense documents are similarly suffused with identity-related statements about ethics and democratic values. The U.K. Ministry of Defence believes it “must be ethical—and be seen to be ethical” in its development of AI.Footnote 95 Further, its AI will be developed for “beneficial” reasons such as “upholding human rights and democratic values.”Footnote 96 As with the United States and Australia, the U.K. constructs itself as ethical compared to its adversaries who “do not adhere to the same moral and ethical frameworks that we do.”Footnote 97 In Australia, one of the leading figures working on ethical AI, Kate Devitt, argued that when it comes to defense AI, Australia should be “a strong, powerful, terrifying goodie” as contrasted with “the baddies.”Footnote 98 Dominant domestic narratives portray Australia as inherently more ethical than its adversaries, with defense industry and former defense science figure Jason Scholz describing its relationship to other countries as an “ethical asymmetry.”Footnote 99 As demonstrated with a small handful of examples here, across Western defense documents and statements from defense-related individuals there is a construction of Western states as inherently ethical actors. It is taken as a given that due to this ethical identity any technology developed and used by these actors will be ethical, and any war fought will be ethical in both justification and means.
The significance of ethics is such that it forms its own economy, with the appeal of the language of ethics obvious enough that defense industry and tech companies have used it in marketing their products. This phenomenon has been described as “economies of virtue.”Footnote 100 The virtue economy is evident in the way in which the defense industry and big tech market themselves toward government defense departments using the language of ethical product design—whether because they actually believe in the ethics, or are using it cynically for branding purposes (or a combination of the two). Big Tech has used ethics narratives to promote its technology for military purposes, painting a picture of humanitarian technology that has better situational awareness than people, tapping into the U.S. military’s desire to “maintain the hegemonic position of the USA as a leader of the democratic forces of the world.”Footnote 101 Google has, for instance, shifted away from the public-facing hesitation to military contracts that it displayed following the fallout of Project Maven, in part through the deployment of an identity-based narrative justifying its involvement in military application of its technology. In February 2025 Google removed a prohibition on using AI in ways that could cause harm from their AI ethics principles, arguing:
There’s a global competition taking place for AI leadership within an increasingly complex geopolitical landscape. We believe democracies should lead in AI development, guided by core values like freedom, equality, and respect for human rights.Footnote 102
We can thus see how ethics is being placed at the forefront of a dramatic rebranding by some companies, from avoiding perceptions of military entanglement to actively advancing military AI. Ethics can then be used as a justification for any actions taken by the actors adopting this technology because it has been linked to the identity of the actor: the technology, and thus the actor, is ethical, good, moral, and humane.
To achieve this branding, companies market themselves on their ethical credentials. IBM, for example, argues that it “leads global efforts to shape the future of responsible AI and ethical AI metrics, standards, and best practices” as it markets itself to defense.Footnote 103 Even Palantir—one of the most controversial companies in this space—has a dedicated and expansive page on its website about the company’s approach to AI ethics.Footnote 104 Companies not only market themselves on their own ethical credentials but also explicitly tie this to identity-related ethical narratives of strategic competition. Palmer Luckey of Anduril, for instance, describes Anduril as “a company that will save Western civilization.”Footnote 105 Anduril aims to lead in “rebooting the arsenal of democracy.”Footnote 106 Luckey sees himself as a paragon of freedom:
Societies have always needed a warrior class that is enthused and excited about enacting violence on others in pursuit of good aims. . . . You need people like me who are sick in that way and who don’t lose any sleep making tools of violence in order to preserve freedom.Footnote 107
This phenomenon exists for companies outside the United States, too. For example, German company Helsing states that it was “founded to put ethics at the core of defence technology development” and that “artificial intelligence will be the key capability to keep liberal democracies from harm.”Footnote 108 Ethics as identity is part of this marketing. Companies market themselves in identity terms, while at the same time “coded ethics becomes a commodity, central to the sales pitch of start-ups.”Footnote 109 The fact that corporate actors use the language of ethics to market their technologies to governments and defense forces is a reflection of the longstanding traditions of how ethics, war, and technology have intersected in the western imaginary, allowing such states to use “ethical” technology that feeds their own sense of moral superiority. As we will see below, ethics as code is another way the interaction of ethics, war, and technology is manifesting in the context of AI and autonomous systems.
Ethics as Code
The idea that ethics can be easily codified and, ideally, quantified, is playing a significant role in shaping Western imaginaries of how AI, machine learning, and autonomous systems can make war more ethical and humane. “Ethics as code” can mean several things, especially in the context of algorithmic warfare. Here I shall focus on approaches to ethics that, broadly speaking, either draw on mathematical, computational, and algorithmic language and reasoning in their approach to ethics or attempt to quantify ethics in some way. This section will explore the pre-existing tendency that western actors have toward an ethics-as-code approach, including the work of roboticist Ronald Arkin as a case study, and cases drawn from DARPA projects that attempt to quantify ethics.
Military ethics has a tendency toward codification. Eastwood argues that we have seen the “reduction of military ethics to the probabilistic management of risks to populations through violence.”Footnote 110 This process of reduction often results in codification. As Elke Schwarz writes:
When the socio-political mandate that informs the norm is centred on a calculable humanity and focused on the abstract idea of the survival of mankind, it is perhaps not surprising that contemporary theories about ethics, specifically in the context of war and just war theory, are turned into something calculable and predictable, framed as formulas or algorithms with which to determine ethical behaviour.Footnote 111
In a military context, ethics is limited due to the inherently militarized nature of the structures it sits within. Ethics tends to be seen “as instrumental and principally related to conduct on and off the battlefield by individual soldiers,” as well as “typically posed as both values to hold and problems to solve.”Footnote 112 Technology has influenced this process, both through changing the way people think and generating a source of hope that quantifiable ethics is possible. Schwarz argues that it is “the logic of computation and information technology” itself that has “become a dominant mode of moral philosophical reasoning in the Anglo-American tradition.”Footnote 113 The ethics of warfare, in such an environment, becomes dominated by “the language of calculable utility and efficiency,” which “tend[s] to draw on a mode of abstract, technical reasoning that mirrors the technologics of the computational processes themselves.”Footnote 114 As Schwarz points out, many of the arguments in favor of using lethal autonomous weapons use mathematical and algorithmic language, employing “analytic hyper-rational reasoning.”Footnote 115 As with the intellectual foundations of machine learning research, here too we find an approach in which rationality, objectivity, and the value of scientific method and technological solutions are foundational. The idea of a codifiable ethics of war is, of course, far from new, but the intellectual inheritances of machine learning reinforce it.
The approach to ethics as code can be seen in military discourse and practice. Weizman relays an example of scientific methods and mathematical logic in relation to Israeli military practices. In a meeting in 2002, a group of military ethicists and legal experts were asked how many civilian deaths they would consider as being in compliance with the principle of proportionality. Their answer was an average of 3.14.Footnote 116 The specific answer is beside the point. The key takeaway from this anecdote is the fact that they give a quantified answer, demonstrating the attempt to quantify an ethical principle of war. Following Schwarz, we can see this as a part of the “scientific-technological mindset” that dominates debates about just war theory in the policy sphere, resulting in a limited approach to ethics in which there is “a prioritisation and privileging of all quantifiable aspects in and of war.”Footnote 117 Included here is what Schwarz calls “a quest for certainty,” encapsulating the belief that there is a fixed and universal ethics to be uncovered.Footnote 118
AI and autonomous weapons systems have been cast as ethical weapons partly due to a belief that ethics can simply be coded into machines, removing the possibility of human error or poor human morality. The goal is to pursue “codes of ethics that instill the good, the lawful, or the normal into the algorithm.”Footnote 119 Ronald Arkin’s work exemplifies this perspective. In his early, influential U.S. defense–funded research, Arkin argued that robots could “potentially perform more ethically in the battlefield than humans are capable of doing.”Footnote 120 He argued they would not have a sense of self-preservation, they would be technologically superior in terms of situational awareness, and they would not be influenced by emotion.Footnote 121 Arkin laid out mathematical formulae as bases for the ethical use of autonomous systems, including constraints drawn from the laws of war and rules of engagement.Footnote 122 One example of his approach is the algorithm that he developed alongside colleagues to “Calculate_Proportionality (Target, Military Necessity, Setting),” which includes instructions such as “Calculate Carnage for the position.”Footnote 123 Critics have described his approach as a “hyper-rationalistic view of encoded ethics.”Footnote 124 The encoding of this algorithm makes an endless series of assumptions about what is ethical and how quantifiable any given ethical principle is. Of course, the principles he discussed—drawn from the laws of war—include demonstrably subjective measures such as military necessity and proportionality. The example of the Bugsplat software discussed earlier shows how these principles would work in an algorithmic context: in that case proportionality was fixed at thirty civilian casualties no matter the military importance of the target.Footnote 125 Another example of adapting the laws of war to new technologies can be found in the approach of Stephen Bornstein, who heads up an Australian defense industry company called Athena AI. He describes how they drafted seventy pages of a legal and ethical framework based on “a variety of ethical principles,” the laws of armed conflict, and the Australian Defense Force’s targeting steps in order to “put it into” their AI product.Footnote 126 Members of the Athena AI project argue that they “have successfully generated a genuine ethics-focused culture and encoded ethical and legal principles within the system.”Footnote 127
Ethics quantification has been proposed in the civilian domain as well as the military domain. Maria Pokholkova and colleagues, for example, argue that given the preponderance of “AI ethics principles” documents and statements, we should “defin[e] quantifiable measures of AI ethics principles.”Footnote 128 This is difficult (or arguably impossible) enough in a civilian setting, and only becomes more difficult (or arguably impossible) in a military setting. One proposed solution to the difficulty of uncertainty and the associated ethical challenges when it comes to AI and machine learning in a military setting is to quantify uncertainty.Footnote 129 Reed proposes that “uncertainty quantification” (UQ) could help build U.S. defense personnel’s trust in AI systems.Footnote 130 The drive to quantify ethics is also evident in the work of DARPA. One project named ASIMOV (the Autonomy Standards and Ideals with Military Operational Values) states its intention “to develop benchmarks to objectively and quantitatively measure the ethical difficulty of future autonomy use cases.”Footnote 131 It aims to respond to one of the key perceived difficulties of ensuring “the ability of autonomous systems to follow human ethical norms.”Footnote 132 DARPA makes it clear that a “quantitative framework” is essential to solving the conundrums of “ethical autonomy.”Footnote 133 In the words of the project lead: “Ethics are hard. Quantifying ethics is even harder … but the stakes are too high not to try everything we can.”Footnote 134 One of the project participants describes this process as “defining a new mathematics of ethics, where ethical scenarios and commander’s intent are represented by knowledge graphs.”Footnote 135 Another project participant describes it in terms of shifting from ethics “at a level of generality” to “a way of evaluating the specific ethical readiness of particular systems for the battlefield.”Footnote 136 Another DARPA project is the Human-AI Communication for Deontic Reasoning Devops program (CODORD). This particular project aims to “efficiently convey knowledge about human intent, laws, policies, and norms into logical programming languages, which an AI can understand.”Footnote 137 In this way, the logic of ethics as code treats ethics as something fixed and achievable—objectively identifiable, and universally good and beneficial for both western societies and the world, and something that exists outside the context in which the relevant weapons or targeting systems will be used. This sits in tension with the politics and lived experiences of military violence as enacted by Western defense forces.
On the Need to Start Foregrounding Experience and Resisting Depoliticized and Decontextualized Ethics
The treatment of ethics as something rational, universal, and objective in both machine learning research and western military thought has several implications. One key effect of a rational, certain, and universal approach to ethics is a depoliticization and neutralization of ethics. To pretend that there is a universal ethics waiting to be uncovered and coded is to see ethics as separated above and beyond politics. It remains largely abstract as it is applied across contexts. By contrast, thinkers such as Zehfuss draw on the work of Derrida to argue that ethics cannot exist separate from politics.Footnote 138 Eastwood further reminds us that ethics is simply “a domain of human activity like any other, rather than . . . the necessarily enlightened pursuit that it is often taken to be.”Footnote 139 One cannot simply take the label of ethics at face value, but must take account of the politics involved in setting the terms of ethical debates and in the application of an ethical label to a particular military practice or technology.
One of the main problems with this depoliticized and neutralized form of ethics is that it leads not to constraining war but rather to expanding and enabling it.Footnote 140 After all, military ethics are always, by definition, “already subordinated to martial violence in that they are always concerned with enabling its infliction.”Footnote 141 They cannot escape the structural confines of the inherently militarized purpose of military forces. It is therefore important to foreground the politics of ethics as Zehfuss encourages us to do. Ethics is too easily used as a commodity and as a justification for war,Footnote 142 feeding into a construction of “virtuous war” and able to create a smokescreen of “technological and ethical superiority.”Footnote 143
One way to foreground the politics of ethics is to bring ethics and the experience of war into conversation. Both ethics as code and ethics as identity remove ethics from the everyday, situated, contextual experience of war, as well as the political realities of decisions to go to war and the political economies underlying those decisions and encouraging the production of endless militarism. In terms of the experience of war, more “precise” technology has resulted in what James Der Derian terms “virtuous war,” or “the technical capability and ethical imperative to threaten and, if necessary, actualize violence from a distance.”Footnote 144 War is sanitized, with the removal of the purveyors of war from the effects of that war. The experience of war by those subject to war tends to be overlooked.
Looking to the experience of living under conditions of a “virtuous” or “ethical” or “humane” war shows that the fixation on the ethics of the means of war misses the reality of war itself. First, there is the way the very idea of ethical or humane war has functioned to expand where and when war is seen as permissible. Second, there is the question of outcomes. Despite the fixation on precision and targeting, the belief in these terms and the standards set as a consequence do not inherently result in fewer casualties or less destruction. One 2009 report, for instance, pointed out that in Pakistan U.S. strikes had killed fifty civilians per militant.Footnote 145 Note, too, the thirty-civilian threshold set for Bugsplat,Footnote 146 or the way Israel simply changed its thresholds for both expected casualties and significance of target.Footnote 147 The idea of civilian deaths as accidental or as an exception has been critiqued before,Footnote 148 but the critique bears repeating. When a set quantity of collateral damage is pre-programmed into code as being acceptable, then those deaths are not accidental. This is heightened in an environment where the combatant-civilian distinction is crumblingFootnote 149—with both the United States and Israel, for instance, tending to consider any male of a certain age as a combatant.Footnote 150 Third, there is the experience of those being subject to “ethical” war. The role of drones, for instance, is one of “terror.”Footnote 151 Their presence creates an environment of constant terror with profound psychological impacts on civilians.Footnote 152 These horrors can only be seen as “ethical” or “humane” war if the experiences of those subjected to them are removed from the equation—whether through a lack of attention or care, a belief in the construction of the West as the ethical actor, or a colonial and racialized excision of certain lives from the narrative of war.
The companion of foregrounding the experience of those subject to war is to foreground the experience of how war is made. By this, I do not mean the more obvious, literal, and immediate experience of soldiers (although this too should be part of the picture)Footnote 153 but rather the broader material realities of how war is made—namely the political economies and the politics of expertise that create the conditions for war to be waged. The narratives, discourses, and political economies that lead to the creation of “ethical” weapons are an essential component of the experience of war. These things create the conditions in which soldiers, governments, publics, and economies will support and be capable of pursuing violence. As we have seen in this article, the language of ethics is a key element of these narratives, discourses, and political economies, mobilized by individuals, companies, and governments in the pursuit of particular interests. Weapons manufacturers are speaking to states and publics using the language of ethics, drawing on both ethics as code (ethics as a simple and solvable problem via mathematical, algorithmic, and computational logic) and ethics as identity (we will help you wage a moral and humane war) to maximize adoption of their technology. Schwarz argues that those who draw on applied ethics are often on a “quest for certainty”:Footnote 154 ethics as code and ethics as identity function to help provide this ethical justification and a sense of certainty to Western warmakers.
It is worth noting that it is not just companies drawing on and creating these narratives, but also experts. Though much more could be written on this, I would briefly divide the experts that drive these narratives into three categories. First, there are the technical, scientific, and legal experts who have been integral to developing technology and assessing the legality of particular strikes. These are the kinds of experts embedded closely within the military domain—defense scientists, military lawyers, and academics collaborating on such projects. Second, there are those who contribute to narratives of geopolitical competition, reifying certain categories of “enemy,” and calling for technological developments to respond to geopolitical uncertainty. These experts include those such as academics or think tankers who argue, for instance, that Western states must pursue AI in order to maintain technological supremacy over China. And finally, given the nature of the technologies under discussion here, an important third set of experts is the messy network of people involved in Big Tech and Silicon Valley—including those driving the AI hype cycle.
Understanding both the horrors of war as experienced by those subject to it and the experience of those creating the conditions for and of war highlights the gap between the two: the ideas of ethical and humane war are coopted by political and economic interests, deeply embedded in capital. This is not reconcilable with the experience of those subject to war, nor the actual political realities of when, how, where, and why Western war has been waged in recent decades. Both forms of experience show that war ethics is not rational, certain, or universalizable but rather constructed, particular, and political.
Conclusion
This article has shown how machine learning, AI, and autonomous systems draw on the Cartesian and Newtonian ideas of rationality, universality, and certainty to construct a narrative of technological solutions to problems of ethics and war. Ethics as code and ethics as identity reify the belief that war and politics can be divorced, and that autonomous systems, machine learning, and AI are the latest solution to exerting technological control over the fog of war. Controlling chaos,Footnote 155 creating a transparent battlefield,Footnote 156 and replacing the “ethico-political dilemmas of war” with “a computational techno-ethics”Footnote 157 sit at the heart of this project. Ethics has been treated as “a methodology” or “a tool,” both within data science and within military thought. It is seen as universalizable and capable of neutralizing power relations.
The article has further shown that claims that certain technologies make war more ethical should be met with high levels of skepticism, for several reasons. The focus on embedding ethics within specific weapons systems narrows dramatically the scope of ethical discussion. Rather than focus on whether war is ethical to wage, ethics as code and ethics as identity in combination mean that, by definition, any war that we-the-ethical wage will be, by definition, ethical because we are who we are, and because we have encoded ethics into our technological systems. The recent history of Western warfare reveals the fallacies involved here: illegal wars, questionably legal wars, the “everywhere war,”Footnote 158 and killing without any declaration of war at all. Israel’s recent use of AI pushes this further, with AI being used to broaden the scope of targeting further and further into the civilian sphere, contributing to “the genocidal acts … committed with intent to destroy, in whole or in part, Palestinians in the Gaza Strip as a group.”Footnote 159
As Birhane argues in relation to data science, ethics should be something that actually alters the way things are done.Footnote 160 What starting to put the ethics of war in conversation with the experience of war tells us is that trying to tame war with ethics or technological superiority is insufficient, as it always has been, and it is driven by people with political and economic stakes in the project. Technology is not a silver bullet able to solve war’s ethical conundrums: war itself is the ethical conundrum. Attempts to tame war therefore require a rejection of the concept of ethical war as enabled by technology and an embrace of ethical perspectives that contain an anti-war foundation.
Open access
