The Politics of Using AI in Public Policy: Evidence from a Field Experiment

• In Event: Poster Session: Innovations in Public Administration Research & Experimental Research
  In Poster Session: Poster Session: Experimental Research

Abstract

Government agencies are routinely using AI-based algorithmic systems to make high-stakes policy decisions (e.g., policing, welfare, education) without the knowledge or consent of the public. This is coming at a significant cost, as the deployment of AI has sparked substantial backlash in several high-profile cases where its usage came to light. How do people view the use of AI algorithms in determining important policies? We explore this question, focusing on the way views evolve in response to personal experience with AI-guided decision making and to exposure to information about the potential implications. We do so using a field experiment conducted in a real-world online labor marketplace, combined with a three-wave panel survey. We track the preferences of over 3,500 workers hired to perform a range of tasks, where the identity of the employer who assigns them to tasks, the content and valence of the tasks they perform are randomly assigned. The results of the experiment will provide insight on the formation and likely trajectory of citizens’ attitudes on AI-usage in shaping public policies.

Authors

• Shir Raviv, Columbia University

• Yotam Margalit, Tel Aviv University