Mediating under the shadow of AI: Public reactions to procedural and substantive roles of AI in court-annexed mediation

Abstract

This study investigates public responses to artificial intelligence (AI) in court-annexed mediation—a setting in which disputing parties retain decision-making authority, unlike the adjudication contexts examined in prior research. Drawing on procedural justice theory, the study develops a role-based framework distinguishing between AI's procedural function (chatbot mediator) and substantive function (AI-generated settlement proposal), and tests how each role shapes fairness perceptions, perceived accuracy, and behavioral outcomes through distinct perceptual pathways. Using a web-based experiment with 1000 participants, the study also examines whether the effectiveness of transparency mechanisms depends on whether explanation types align with the role the actor plays in the mediation process. Results indicate that initial public skepticism toward AI mediation is significant but attenuates after participants engage with the process and review the settlement proposal (mixed-effects: drafted-by-AI × stage 1 ≈ 0.40; chatbot × stage 1 ≈ 0.24). AI-generated plans were judged significantly fairer (b ≈ 0.25, p < 0.05) and more accurate (b ≈ 0.22, p < 0.05) than human-drafted ones, and these perceptions mediated acceptance and satisfaction. Transparency strategies produced asymmetric effects: global (rule-level) explanations improved accuracy and acceptance only when paired with procedural AI, whereas local (case-specific) explanations backfired when combined with human-authored proposals. These findings suggest that integrating AI into mediation can enhance both efficiency and perceived legitimacy—provided that parties maintain ultimate decision control and that explanation types are matched to the AI's procedural or substantive role.