Framing responsibility: Human and AI agent effects on apology effectiveness in service failures

Abstract

As artificial intelligence (AI) systems become increasingly prevalent in service interactions, understanding how people assign responsibility and respond to apologies from AI versus human agents is critical for designing effective communication strategies. This research examines how the type of service agent (human vs. AI), the nature of a crisis (value-based vs. performance-based), and attribution strategy (internal vs. external) jointly shape individuals’ perceptions and evaluations of crisis responses. Across two experimental studies, we show that people interpret the moral and functional accountability of agents differently depending on the type of failure and the perceived capacity of the agent. In Study 1, value-based crises elicited stronger negative reactions when a human agent was involved, whereas AI agents were evaluated more harshly in performance-based failures. Study 2 introduces attribution strategy as a moderator and reveals that the effectiveness of an apology hinges on the congruence between agent type, crisis type, and attribution framing. Internal attributions were more effective for human agents in value-related crises and for chatbot agents in performance-related ones, while external attributions were more acceptable in contexts where the agent was not perceived to bear moral or functional responsibility. These findings apply attribution theory to the context of AI-mediated service crises by highlighting agent–crisis–attribution fit as a key determinant of apology effectiveness, with implications for apology design, organizational accountability, and the future of human-machine communication in digital service environments.