No Mere Deodands: Human Responsibilities in the Use of Violent Intelligent Systems Under Public International Law

Abstract

The tide toward the militarization of autonomous technologies has prompted critics to propose a pre-emptive ban on their development for fear that they may not adhere to international laws and, worse still, that no one will be responsible for their use. These criticisms, however, are rooted in pessimistic prognoses that misconstrue the potential of emerging technologies and international law’s ability to regulate them. Accordingly, this thesis advances three arguments to dismantle these dystopian perspectives. First, a pre-emptive ban ignores the centuries-long distribution of violent tasks between humans and non-human actants. In the process, it seeks to revise current terminologies by shifting the focus on autonomy and lethality toward intelligence, violence, and systems. Second, the current international legal architecture is adequate to (a) ensure the responsible use of emerging autonomous weapons systems and (b) allocate human responsibility for their use. Critics often argue that it is impossible to preprogram all eventualities of warfare into a machine, but this perspective ignores advances in machine learning, that enable intelligent systems to teach themselves rules based on set parameters and algorithms. Third, critics misunderstand the networked nature of human violence, and consequently underestimate the elasticity of international law. To this end, this thesis borrows from evolutionary biology, psychology, and semiotics to explain the composition and constitution of networks of human violence. Ultimately, by viewing armed conflicts through the lenses of networks, this thesis argues that international law is capable of regulating human violence regardless of its conduit.