When AI feels unobserved: consumer unethical behavior in AI-mediated post-Service

Abstract

As AI service agents are increasingly deployed in post-service interactions such as complaints and compensation requests, concerns have arisen about consumers’ unethical behavior toward AI. While prior research has largely focused on emotional or anthropomorphic explanations, this study examines cognitive and contextual mechanisms underlying unethical behavior in AI-mediated post-service contexts. Drawing on Social Presence Theory, we propose that interacting with AI (vs. human) service agents reduces perceived social presence and monitoring, lowering moral constraints and increasing unethical behavior. Across three experiments, we find consistent support for this mechanism. Moreover, AI self-efficacy serves as a critical individual-level boundary condition. Higher AI self-efficacy strengthens unethical behavior by fostering an instrumental perception of AI agents as controllable tools rather than social entities, further weakening perceived observability. This research extends social presence theory to moral decision-making in AI-mediated services, reveals the dark side of AI self-efficacy, and offers actionable insights for managing post-service.