Navigating Institutional Complexity: How Competing Institutional Logics Shape AI Adoption in Local Governments

• In Event: AI Adoption and Institutional Governance

Abstract

Despite the rapid spread of artificial intelligence (AI) systems in the public sector, local government adoption remains highly uneven. While prior research predominantly attributes this variation to insufficient technological capacity or organizational resources, this study moves beyond traditional capacity-based explanations, and argues that organizational responses to emerging technologies are structured by deeply ingrained, competing institutional logics. Drawing on institutional logics theory, this study adopts a configurational approach to investigate how the institutional complexity of bureaucratic, market, and participatory logics shapes AI adoption in local governments, focusing specifically on generative AI tools for administrative tasks and AI-enabled chatbots deployed on government websites. Using a 2024 survey of public managers from 303 local governments in the U.S., this study employs latent profile analysis (LPA) and a two-stage hurdle regression model. LPA clusters local governments into three profiles: bureaucracy-oriented, market-oriented, and contested profile where multiple logics exhibit similar salience. Results indicate that the influence of these profiles is contingent upon the technology’s innovation pathway. Governments in contested profiles are significantly less likely to adopt generative AI tools, which often emerge through bottom-up, employee-driven processes. In contrast, the adoption of AI chatbots, typically introduced through top-down mandates. exhibits configurational equifinality, with similar adoption rates across profiles. Furthermore, the two-stage hurdle regression model shows that the effects of each institutional logic on the adoption of generative AI tools are dependent on organizational configurations. Bureaucratic logic is negatively associated with adoption in contested configurations but becomes positive in market-oriented contexts. Market logic is positively associated with adoption in organizations dominated by a single logic but loses significance under contested configuration. Participatory logic shows limited relevance to adoption across configurations. Finally, while institutional logics play a prominent role in shaping governments’ initial adoption decisions, they have a limited role on the subsequent extent of AI use.Taken together, these findings advance understanding of how institutional complexity and innovation pathways jointly shapes public organizations’ situated responses to AI. They also carry practical implications for public managers navigating the adoption and scaling of AI technologies.

Author

• Jinghuan Ma, Arizona State University
  Presenting Author