Autonomous Weapons Systems and International Norms

  • Published

Autonomous Weapons Systems and International Norms by Ingvild Bode and Hendrik Huelss. McGill-Queen’s University Press, 2022, 281 pp.

In Autonomous Weapons Systems and International Norms, authors Ingvild Bode and Hendrik Huelss make a novel contribution to the literature on autonomous weapons systems (AWS). The authors, both professors of international politics at the University of Southern Denmark’s Center for War Studies, situate the discussion of AWS development and usage within the context of norm-making in international relations. The main argument of the book is that international law is indeterminate in the formation of norms regarding the use of force and that since AWS development and use have so far outpaced international regulation, such norms are more likely to be shaped by practice.

Chapter 1 provides an excellent overview of international lawmaking discussions regarding AWS to date, primarily drawn from meetings of the UN Charter for Certain Conventional Weapons Systems. The regulatory proceedings are beset by the problem of definitions from the start, such as the difference between autonomy and automation and what constitutes “meaningful human control” to what extent artificial intelligence (AI) and machine learning are integrated into these systems, and accounting for autonomous systems with dual use capability (lethal and non-lethal). The authors provide a valuable roadmap for the terminologies and classifications of these technologies while pointing out that international regulations have fallen far behind the pace of AWS innovation and their employment by developed countries.

In chapter 2, the authors discuss historical case studies regarding the emergence of norms for four technologies of war: submarines, chemical weapons, nuclear weapons, and blinding lasers. The overall finding is that international law has been indeterminate in shaping the combat employment of these weapons systems. Chapter 3 examines how drone technology has influenced the norms surrounding the use of force, specifically attribution, imminence, and targeted killing. Chapters 3 and 4 then argue not only that the “logic of appropriateness” behind the use of force is shaped by international humanitarian law and international humanitarian rights law, but also that the legal-public sphere of international law (macro level) interacts with a procedural-organizational sphere in a way that gives rise to the novel concept of procedural norms (135). Procedural norms are “not the outcome of formalised, public, deliberative negotiations” (149). Instead, they emerge from what is considered useful and efficient by organizations such as the military and corporations at the meso and micro levels.

The book’s main arguments regarding international norms thus turn broadly upon literature from the constructivist school in international relations. Despite its important contributions to neorealism and neoliberalism, the constructivist school tends to be hamstrung by the rhetorical density of its arguments, and this book cannot help but suffer many of the same pitfalls. There is also a curious devotion to acronyms and parenthetical citations throughout, which only contributes to the density. While international relations scholars may be fascinated by the exposition on international norm-making here, the lay reader will likely be overwhelmed by the heaviness of the prose.

In chapter 5, the authors examine the emerging norm of meaningful human control in autonomous weapons systems using the example of autonomy in modern air defense systems. Most modern air defense systems can operate in manual and automated modes, with various levels of autonomy ranging from human-in-the-loop to full autonomy. The authors use five case studies of failures and fratricide in modern air defense systems to argue that due to emerging threats and the fallible nature of human-machine interaction, the notion of human control over these systems has become “meaningless” (193). Here, the authors’ analysis contains several problems.

First, the case studies chosen by the authors all exhibit human failure as a root cause, so the connection of the failure to autonomy is weak. Indeed, some human failures give the arguments of AWS proponents—that AWS could mitigate human failings and produce less suffering and collateral damage— added weight. Second, the idea that full autonomy exists in modern air defense systems is debatable. The authors’ preferred definition of autonomy, namely “the ability of a machine to perform a task without human input,” contradicts their claims of true autonomy in modern air defense systems, since human input is present in the design and programming of these systems before they are ever placed into automatic modes (19). In the case of Aegis specifically, the system in any automatic mode will only fight and conduct engagements exactly as humans prescribe it to, and the decision to enter automatic mode can be revoked at any time. Furthermore, the system is not a black box like AI or a machine-learning algorithm. We can ask the system to recall why it performed the way it did, and its answers are unambiguous.

The authors’ fixation on air defense systems as a cautionary example of the slippery slope of decreasing human control does not meet with the reality of how these systems are employed, and the five failures cited—most of which occurred in the fog of combat conditions—ignore the much more numerous successes. If there is a slippery slope in human control, there has been ample opportunity for the legal-public sphere to bring that to light in the intervening years. That this has not occurred does not warrant the hyperbolic assertion that “the integration of autonomous features into air defence systems has not been questioned” (204).

In chapter 5 and the ensuing conclusion, the authors editorialize about the kinds of norms they would prefer to see and imply that procedural norms are somehow less legitimate than deliberative ones. Some value judgments seem out of place given the book’s central argument: “The fact that a machine, rather than a human, fails matters. It matters because handing over life and death decisions to machines violates human dignity by reducing human beings to objects” (205). Here, the authors decry public advocacy focused on technological solutions to these life-or-death problems; however, situated within the authors’ own thesis, it seems that “technological solutionism” is actually a procedural norm that aims to ameliorate the fallible nature of human-machine interaction faster than the legal-public sphere allows (206).

Throughout the reading, the authors attempt to move the discussion of norms beyond the realms of international legal frameworks, but legal frameworks such as UN charters and International Court of Justice rulings are used as a crutch. This is likely because these deliberations constitute the chief means by which implicit international norms concerning AWS have been rendered explicit. Still, the overwhelming focus on legal deliberations on the AWS issue obscures many of the asserted linkages between legal-public and procedural-organizational spheres and muddies the central argument.

In the conclusion, the authors lament the validation of their original hypothesis and call for a decoupling in the other direction, in favor of deliberative norms to stop the dilution of meaningful human control in AWS development. This strident advocacy mostly distracts from the book’s main argument and would carry more weight if the authors had successfully demonstrated the dangers of non-deliberative processes. By resorting to this advocacy, the authors risk contributing to the alarmism they decry when dismissing science-fiction portrayals of autonomous machines in chapter 1, bringing heightened emotion into an otherwise sober-minded exposition of AWS and international norms.

Overall, this book succeeds in illuminating the intersection of international relations with moral, legal, and normative frameworks regarding the development and use of AWS. Its novel contribution illustrates the interplay of procedural and deliberative norms and how this interplay shapes emerging norms concerning lethal AWS. The extensive constructivist discussion on legal-public spheres versus procedural-organizational spheres, deliberative versus procedural norm-making, positivist versus post-positivist frameworks, agent versus structure, micro versus macro, and normativity versus normality all contribute to the central idea of procedural norms, but this structure stands like a Jenga tower: it wobbles the more logs are added or removed.

International relations scholars will gain the most from this discussion, although policymakers and public policy advocates may find the early chapter overviews and final recommendations useful. Yet the authors’ approach would benefit from a broader look at the emergence of norms surrounding autonomous machines and AI as a whole, given that the slippery slope, if there is one, has the potential to impact not just the use of force, but all facets of human life.

Commander Jason R. Highley, USN

"The views expressed are those of the author(s) and do not reflect the official policy or position of the US government or the Department of Defense."