Framing artificial intelligence as a policy instrument in urban climate action: Practitioners' perspectives from Paris

Urban practitioners increasingly integrate artificial intelligence (AI) into tasks related to the implementation of climate action. With the impact of AI transcending specific tasks but changing responsibilities and routines, it must be considered as a policy instrument. While research identified potential AI applications and traced implementation processes, it remains understudied how practitioners introduce AI and how AI systems change workflows and decision hierarchies in urban climate action. Treating AI as a policy instrument, this study examines the city of Paris through 10 open interviews with 12 interviewees and 13 documents. We show that AI currently produces incremental shifts in tasks related to climate action. These shifts redefine urban practitioners' responsibilities towards integrating AI based on socio-technical risk assessment and translation to deliberative formats. We also document how practitioners mediate AI's integration into urban climate action through technocratic, moral, and political logics of action, contrasted with their doubts about the usefulness of AI. Doubts include, for instance, that AI risks depoliticizing urban climate action. By assessing AI as a socio-technical policy instrument rather than a neutral tool, our study unpacks both technical operational utility and symbolic values, including tensions in between, such as complexity-handling versus oversimplification, and acceleration versus accountability. We argue that urban practitioners must critically reflect on their expectations towards AI systems and engage with fundamental concerns to reduce tensions. Finally, we created a checklist for urban practitioners who engage with AI in climate action and aim to operationalize democratic AI governance as a means to reduce adverse consequences.