The accelerating arms race in Artificial Intelligence and the diffusion of cheap, technologically advanced military systems among state and non-state actors, compel countries to adopt robotic and autonomous systems (RAS). This is due to not only the prospect of lagging behind allies, but the notion that adversaries use RAS to gain a significant military advantage and escalation dominance. How can countries stay competitive, without violating core national and international ethical principles?
This paper discusses the key ethical challenges spanning human agency, human dignity and responsibility in the operation of RAS. It presents a three-part framework for how to view human control over RAS in an ethical manner:
- Life cycle perspective
- Sub-system functionality perspective
- Observe-orient-decide-act (OODA) loop perspective
Based on the framework above, the paper presents the following concrete recommendations for actors engaging in the design, development, testing, operation and decommissioning of RAS at the strategic and operational levels:
- The fundamental principle is to work with ‘ethics by design’: ethical considerations are incorporated in not just the operating of RAS, but all throughout a system’s life cycle;
- Create an institutional culture of shared accountability concerning all actors involved throughout the RAS life cycle;
- Build a wider understanding of system behavior by involving end users as early as in the design and testing stages. This way, the users within the military can trace, understand and predict a system’s decision-making process;
- Develop best practice guidelines for (1) outsourcing to external contractors; and (2) interoperability with technologically advanced allied armed forces that co-deploy RAS;
- Identify within what sub-system functions of RAS increasing automation and autonomy will present benefits to the military without eliciting major ethical concerns, such as movement controls and computer vision;
- Embed core rules of engagement (ROEs), along with an open architecture that allows for mission-specific ROEs by mission command;
- Improve transparency on the uses and contexts of use of RAS in the military domain with the general public and inform policymakers of realistic opportunities and limitations of complex system design.
Dowload the report here.
Responsibility for the content and for the opinions expressed rests solely with the authors. Publication does not constitute an endorsement by the Netherlands Ministry of Defence.