Ethical Considerations for AI in Local Government Human Resource Management

• In Event: Public Administration in a Local Government Context

Abstract

Prior work on artificial intelligence (AI) policy has primarily focused on the national and state levels, with some growing attention to the local level (Busuioc 2021, Horowitz and Kahn 2021, Liebig et al. 2022, Yigitcanlar et al. 2022). However, less is known about the design and implementation of AI tools at the local level, especially for the operation of local government itself. AI applications are increasingly used for employee monitoring and for job selection and promotion, and recent regulations such as New York City’s Automated Employment Decision Tools Law increase scrutiny of these tools. Drawing on a sample of over 10,000 individuals who have served as the Head of HR in municipal or county government since 2021 across the United States, we field a survey to understand 1) how automated decision systems are currently used for HR purposes in local government, 2) the perspectives of HR administrators regarding how these tools should be used, and 3) which values and considerations they deem most important when deciding whether to adopt or implement a tool. Additionally, we include an embedded conjoint experiment to explore how local government HR administrators evaluate the relative importance of model performance, vendor transparency, fairness across demographic groups, explainability for administrators, consent and transparency for impacted individuals, data scope, human oversight, and cost. We also plan to compare this public sector sample to a private sector sample of HR managers. The results have important implications for the regulation of AI, for the operation of local government in the AI era, and for the selection of government officials who become the "face of the state" as street-level bureaucrats and who create (or not) representative bureaucracy.

Authors

• Daniel S. Schiff, Purdue University

• Ian Adams, University of utah

• Kaylyn Jackson Schiff, Purdue University

• Josh McCrain, University of Utah