Artificial Intelligence for Better Protection of Civilians During Urban Warfare
Editor’s Note: This post is drawn from the author’s article-length work recently published in The Military Law and the Law of War Review.
Images emerging from hostilities in cities such as Gaza, Odesa, Kharkiv, and Donetsk send a stark reminder that warfare causes relentless suffering and devastating consequences for people’s lives and livelihoods. At the same time, the events reveal the urban warfare challenge militaries face in implementing international humanitarian law (IHL) rules and principles on the conduct of hostilities to effectively protect civilians. This challenge is largely due to the physical structures of cities, population density, and civilian infrastructure.
Given these practical challenges, new approaches to reducing the humanitarian consequences of urban warfare must be placed at the fore. Indeed, there is a growing realization that this can be achieved through measures that mitigate the risk of harm to civilians well beyond assessing the lawfulness of an attack.
As a general-purpose technology, artificial intelligence (AI) offers great potential to enhance a wide range of military activities, including capabilities to mitigate civilian harm during urban warfare. However, as I have explored here, this has not been the focus of governments thus far. They have preferred to invest in faster and more precise targeting decisions rather than the protection of civilians more broadly. It is in response to this regretful observation that I wish to highlight in this post how, despite the risks associated with the use of AI systems, these technologies offer militaries enormous opportunities to improve their civilian harm mitigation capabilities.
Recognizing the Value of Civilian Harm Mitigation Measures
Civilian harm mitigation measures essentially consist of efforts by belligerents to avoid, minimize, and mitigate the harms that may accompany their operations. Indeed, in recent years, States and organizations have adopted policies and guidelines that establish measures in this regard and recognize the protection of civiliansas a “strategic and moral” imperative in all types of warfare. As outlined in the U.S. Department of Defense (DoD) Civilian Harm Mitigation and Response Action Plan, “[p]rotecting civilians from harm in connection with military operations is not only a moral imperative, it is also critical to achieving long-term success on the battlefield.”
These shifts attest to an increasing recognition of the need to adopt mitigation measures beyond the moment of attack, including in the training and planning phases, all the way through to after the last shot is fired, for example by investigating civilian harm. As explained in the European Union Military Staff’s (EUMS) Concept on Protection of Civilians and EU-led Military Operations, the protection of civilians needs “to be addressed through the whole spectrum of military activities: education, training, planning, conducting operations, reviewing and lessons learned.” In other words, effective protection of civilians requires that militaries identify opportunities to prevent, reduce, and mitigate civilian harm at all levels of planning, i.e. strategic, operational, and tactical levels. These are welcome developments, given the increase in conflicts in urban areas where the protection of civilians during military operations has become one of the major challenges for the warring parties.
AI to Enhance Civilian Harm Mitigation Measures
AI systems can be developed and used to enhance civilian harm mitigation measures across all military planning phases. In the pre-deployment phase, AI systems could contribute to the training of militaries for urban operations. In particular, urban military training could be complemented by virtual simulations that better reflect the multidimensional structure of cities as well as the civilian dimension of operations in urban environments. In addition, AI could be used to take on the role of adversaries to support the development and improvement of operational concepts, tactics, and strategies. Ultimately, this would not only provide militaries with a better understanding, but also new solutions in response to the enemy’s tactics that endanger the civilian population and their livelihoods.
In the planning phase, AI systems can help the military better understand the civilian dimension of the terrain by mapping critical civilian infrastructures, such as hospitals, power plants, and water sources. These systems could also be used to map the interconnectedness of these infrastructures, a crucial step in reducing the impact on civilians and their livelihoods and allocating resources efficiently when damage occurs. As seen in Ukraine, electric power plants have routinely come under attack, “cutting electricity, heat and hot water in many cities and forcing factories in some areas to temporarily close.” Importantly, while some of these infrastructures can be restored within a few days or weeks, others must be rebuilt from scratch.
At the operational level, AI systems can conduct civilian patterns of life analysis based on different sources of intelligence more effectively than human data analysts. These assessments offer an “understanding of civilian behavior and movement within and in/out of an urban area, which will lead to a holistic understanding of potential risks of civilian harm.”
At the tactical level, AI systems may soon be able to help militaries overcome the difficulties of operating inside and between buildings by enabling them to see through opaque structures, such as walls, without physically entering a room. By reducing the uncertainty of locating military objectives, the risk of falsely targeting civilians inside buildings is also reduced. In addition to faster and more accurate targeting decisions, continued advances in AI systems for object recognition can further improve militaries’ ability to protect civilians from harm. For example, a report by the Center for Naval Analysis describes the potential of AI systems to warn of the presence of fleeing civilians, detect potential miscorrelations, identify a change in collateral damage estimate, or recognize protected symbols.
The effective mitigation of civilian harm also requires post-attack actions such as after-the-fact assessments of military operations to inform the development of doctrine, future planning, and conduct of operations. In this context, AI systems have the potential to facilitate not only processing of data on civilian casualties, but also gathering and analysis of information on the tactics employed by militaries in urban areas. In this way, AI systems can assist militaries to better understand which civilian harm mitigation measures they employ, when to apply them, which measures were successful in which situation, which were not, and why.
Lastly, once hostilities have ended, AI systems can support militaries to mitigate civilian harm, for example in the context of antipersonnel mine clearance operations. By automatically detecting landmines using thermal imaging, AI systems could help militaries in detecting mines more quickly and accurately.
Of course, like any other use of AI systems, some of these propositions come with associated risks. While my focus here is on the opportunities of these systems, it is important to underscore that any proposed AI application requires careful consideration of the risks against the benefits for the protection of civilians.
Final Thoughts
As unfortunate as it is, we cannot ignore the fact that cities and other urban areas are likely to remain the main battlegrounds in future conflicts. Recent images of civilian suffering and death in conflict-affected cities around the world should prompt governments to adopt new means to alleviate this suffering.
In this post, I have only scratched the surface of the enormous possibilities that AI systems could offer militaries for this purpose. Nevertheless, I believe I have covered enough ground to show how AI can be an important means of improving the military’s ability to alleviate the suffering of civilians in urban warfare. Much will now depend on whether governments decide to invest in developing these AI potentials. These are not only within the realm of the possible, they are also urgently needed.
***
Anna Rosalie Greipl is a Research Assistant at the Academy of International Humanitarian Law and Human Rights (Geneva Academy), where she works for the Digitalization of Armed Conflict projects.
Photo credit: Sgt. Kassie McDole
