Description
AI decision-support systems are entering nuclear command structures with little understanding of how they will fundamentally alter crisis decision-making and human control over weapons of mass destruction. How will traditionally human dominated nuclear decision-making be affected by AI-DSS integration?
This research establishes a baseline understanding of AI-related escalation risks in this fast-paced, evolving field. Contrary to assumptions that AI improves decision quality, evidence suggests AI-DSS may systematically bias decision-makers toward aggression by accelerating use-of-force decisions and creating illusions of certainty in inherently uncertain crises.
Drawing on behavioral decision-making theory, this paper analyzes how AI integration with nuclear systems could realistically unfold, and builds a framework to assess escalation risks. This theoretical analysis is complemented by comparative examination of alleged real-world AI-DSS deployments in Gaza and Ukraine for target selection and determination.
The research identifies which cognitive risks (e.g., automation bias, complacency, and anthropomorphization) could manifest themselves in the nuclear command structure and pose the greatest danger in engendering aggression and weakening decision-making heuristics. By determining the severity of risks, and which are empirically supported given available data, this paper prioritizes avenues for mitigation efforts and establishes a foundation for future research in this critical domain.
