This paper, co-authored by Tzuhao Chen, Mila Gasco-Hernandez and Ramon Gil Garcia is published in Public Administration at https://onlinelibrary.wiley.com/doi/abs/10.1111/padm.70051

Recent research has examined various aspects of the government’s use of artificial intelligence (AI), including its implications for accountability as well as the policy and management recommendations proposed to address some related concerns. However, two important gaps remain and need further empirical exploration. First, it is unclear why accountability in AI-based system use is viewed as such a critical issue. Second, it is uncertain to what extent existing policy and management recommendations from academia are reflected in government policies. This study aims to bridge these gaps by empirically investigating how the public sector addresses accountability in the use of AI-based systems. Based on an analysis of policy documents related to AI-based systems from 32 U.S. states, our findings show why accountability matters, who is being held accountable to whom, by what standards, and with what consequences from a practical perspective. Furthermore, we identify several research-practice gaps that merit further exploration.