Not Oracles of the Battlefield: Safety Considerations for AI-Based Military Decision Support Systems
Year: 2024
Type: conference-paper
Abstract: AI-based military decision support systems that help commanders observe, orient, decide, and act on the battlefield are highly sought after by military leadership. With the advent of large language models, AI developers have begun advertising automated AI-based decision support systems designed to both analyze and act on data from the battlefield. While the desire to use decision support systems to make better decisions on the battlefield is unsurprising, the responsible deployment of such systems requires a clear understanding of the capabilities and limitations of modern machine learning models. This paper reviews recently proposed uses of AI-enables decision support systems (DSS), provides a simplified framework for considering AI-DSS capabilities and limitations, and recommends practical risk mitigations commanders might employ when operating with an AI-enabled DSS.
Authors: Emelia Probasco, Matthew Burtell, Helen Toner, Tim G. J. Rudner
Institutions: Georgetown University, New York University
Language: English
Subfield: Control and Systems Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W4404518557