Public Perception on Authority, Transparency, and Bias of AI in Courts: A Vignette Experiment with Chat Intervention
Huyen Nguyen (University of Hamburg); Arna Woemmel (University of Hamburg); Hendrik Huening (University of Hamburg); Lydia Mechtenberg (University of Hamburg)
Abstract
How do ordinary citizens perceive the increasing influence of artificial intelligence (AI) in society? What significance do they attribute to an ethical design and application of AI? This study investigates the normative attitudes of 2,864 subjects in the UK toward the application of AI as a decision-support tool in criminal courts, using a vignette experiment. We vary the trade-off between discrimination-free versus informed judgments of AI in courts, its degree of decision authority, and the transparency of its code, to study the normative value individuals attribute to these design features of AI. Our results show that transparency is the only factor with a significant impact on participants' attitudes. After the vignette experiment, we implement a real-time randomized chat among participants to investigate the potential changes in attitudes due to peer arguments. We find that participants become more skeptical about AI usage in courts after chatting with others. This finding highlights the importance of understanding public discourse in shaping attitudes toward AI usage in the public domain.