The Politics of AI Adoption in Public Administration: A Framing Analysis

Artificial intelligence (AI) systems are increasingly being adopted by public administrations worldwide. However, the drivers and implications of AI deployment in the public sector remain uncertain. In particular, it is unclear how the use of AI by public administrations is conceptualized on the ground, whether primarily as a means to enhance legitimacy, efficiency, or innovation, or instead as a source of risks related to bias, accountability, and privacy, both in public discourse and among political-administrative actors. To address this gap, this paper combines a large-scale analysis of textual data (examining tone, sentiment, themes, and framing) with an in-depth comparative case study of two contrasting political-administrative contexts: Estonia and Switzerland. Estonia represents a pioneer in digital government and early AI adoption, while Switzerland, partly due to its highly decentralized federal structure and strong tradition of direct democracy, exhibits a more cautious and fragmented approach to AI implementation in the public sector. By doing so, the paper sheds light on how institutional, organizational and regulatory contexts shaping the framing, and thereby the modality of adoption, of AI in public administrations.