Policy-driven innovation: The science-policy nexus in artificial intelligence research in Germany

This study reconstructs and characterises the science-policy nexus in the field of artificial intelligence research in Germany, examining the alignment between normative policy goals and empirical research outcomes. Adopting a narrative policy framework, the research investigates transition dynamics across policy, scholarly, and innovation levels, tracing the interconnectedness in stages of patents, papers, and publicly funded projects. The research employs a two-pronged empirical approach: 1) identifying German contributions to AI research through patent citations and bibliometric data, and 2) linking these outputs to the policy instances that funded and enabled them. This methodology reveals the complex pathways through which policy intentions translate into research outcomes, highlighting the mostly indirect nature of this relationship. Key findings emphasise significant challenges in data quality and availability, particularly in linking research outputs to higher-level policy dimensions. While the study successfully identifies German-affiliated papers through patent and bibliometric datasets, it uncovers fundamental disconnections between stated policy objectives and actual research trajectories. These dimensions are compounded by administrative bottlenecks, asynchronous implementation settings, and entangled funding periods. The study concludes that AI development is heavily influenced by geopolitical and strategic decisions extending beyond academia, and that academic research into AI is part of a larger narrative of policy and political design. The study offers a framework for assessing the nuances of science-policy pathways while acknowledging the limitations of fully characterising this nexus. The systematisation presented serves as a foundation for future research, emphasising the need for comprehensive and coherent data sources to evaluate the phases and connections within scholarly and policy narratives.