Did that AI just Charge me a Fine? Citizens' Perceptions of AI-based Discretion in Public Administration

Abstract

Automation and Artificial Intelligence (AI) are continuously advancing decision-making in public administrations. Research focuses on investigating benefits and challenges of AI agents that make decisions without human-intervention, namely AI-systems that independently exercise administrative discretion. However, little is known about citizens' perspective of such autonomous systems. To address this gap, we investigated citizens' fairness and satisfaction perceptions of an AI-based self-service kiosk that performs administrative discretion. Through a controlled Wizard of Oz experiment, we assessed how citizens perceived the kiosk's decision to either impose or waive a fine. Precisely, we measured how engaging citizens in the decision-making process through a dialogue affect their perceptions. We contribute novel citizen-oriented views on how to apply AI-based discretion in public administration. Our results revealed that while citizens do not generally oppose delegating discretionary power to fully autonomous systems, engaging them in the decision-making process through a dialogue will positively alter their perceptions. Moreover, we provide insights on how other factors (i.e., decision justification and citizens' self-gain) influence citizens' perception of AI-based administrative discretion.