Multi-layered discourse shaping military AI: the cases of Germany and the UK

Abstract

Artificial intelligence (AI) is being increasingly utilized by militaries across the globe, with major powers like the USA and China leading the way. Indeed, from the perspective of various realist theories, it can be expected that all countries with sufficient resources for developing military AI capabilities will do so. However, there are instances of countries with sufficient resources not showing any substantial military AI practices, defying realist expectations. This study proposes an alternative explanation to realist theories for the differences in the scope of military AI practices by states, arguing that ideational conditions like norms, ethics, and identity are decisive rather than structural pressures. To answer the research question “What explains difference in the scope of military AI practices by states?”, the study formulates a theoretical framework integrating Strategic Culture and Sociotechnical Imaginaries as a country’s deeper discourse layers within Ole Wæver’s multi-layered discourse analysis model. This framework is then applied within a most similar systems design, controlling for realist conditions and selecting Germany and the UK as case studies with differing dominant discourses on military AI. Thereafter, detailed discourse analysis on dominant discourses on military AI is conducted for both cases, and their scope of military AI practices is determined based on the number of military AI applications, expert assessments, and specific instructions, policies, and doctrines for military AI. Germany showed a cautious dominant discourse on military AI and a limited scope of military AI practices, while the UK showed an embracing dominant discourse on military AI and a comprehensive scope of military AI practices. Hence, the discourse- theoretical framework developed in this study offers an explanation superior to realist accounts and contributes to the literature on military and security studies more broadly by offering an innovative approach to studying general enabling technologies. It also has important policy implications for AI arms control and diplomacy.