The contingent effects of AI transparency in service recovery: a moderated mediation model of perceived warmth and competence

Purpose This research investigates how artificial intelligence (AI) transparency shapes consumer responses in service recovery. By revealing the reasoning process behind AI decisions, transparency functions as an anthropomorphic cue that humanizes the system's communication. Drawing on social cognitive theory, the research develops and tests a generalizable mechanism framework linking AI transparency, consumer cognition and recovery satisfaction, while identifying boundary conditions. Design/methodology/approach Three scenario-based experiments were conducted. Study 1 (N = 98) tested the direct effect of AI transparency on recovery satisfaction. Study 2 (N = 230) examined the parallel mediating roles of perceived warmth and competence. Study 3 (N = 284) tested the moderating role of service failure severity and the corresponding conditional indirect effects. Findings AI transparency significantly improves recovery satisfaction, indirectly through perceived warmth and competence. Service failure severity moderated the effects of transparency on both mediators. The indirect effects through warmth and competence were significant under low service failure severity but not under high severity, identifying severity as a boundary condition for the effectiveness of AI transparency in service recovery. Originality/value This research conceptualizes AI transparency as a form of anthropomorphism that reveals AI's thought process, thereby extending the application of social cognitive theory to AI-mediated service recovery. It explains how transparency enhances consumer satisfaction through warmth and competence and clarifies how these effects vary with service failure severity.