Paper Type
Short
Paper Number
PACIS2025-1504
Description
The rapid development of AI has transformed customer service, yet AI-powered chatbots remain prone to service failures. While prior research has examined recovery approaches and tone of voices separately, their combined effects on user perceptions and emotional responses are underexplored. This study adopts the Stimulus-Organism-Response (S-O-R) framework to investigate how tone of voices (formal vs. informal) and recovery approaches (informational vs. emotional) interact to influence user perceptions of AI reliability, trust, and rapport, and how these perceptions in turn shape negative emotional responses. A 2×2 between-subjects experimental design manipulates tone and recovery strategy in chatbot-generated recovery messages. By observing user reactions to controlled AI failure scenarios, the study aims to clarify how communication congruence or mismatch affects cognitive and emotional responses. The findings will provide actionable insights for optimizing chatbot communication, contributing to a comprehensive understanding of AI-mediated service recovery and advancing theoretical and practical knowledge in this emerging area.
Recommended Citation
Yeh, Gary Yu-Ho; Chiu, Chao-Min; and Hsu, Jack, "When AI Apologizes: The Impact of Recovery Strategy and Tone on User Response" (2025). PACIS 2025 Proceedings. 9.
https://aisel.aisnet.org/pacis2025/hci/hci/9
When AI Apologizes: The Impact of Recovery Strategy and Tone on User Response
The rapid development of AI has transformed customer service, yet AI-powered chatbots remain prone to service failures. While prior research has examined recovery approaches and tone of voices separately, their combined effects on user perceptions and emotional responses are underexplored. This study adopts the Stimulus-Organism-Response (S-O-R) framework to investigate how tone of voices (formal vs. informal) and recovery approaches (informational vs. emotional) interact to influence user perceptions of AI reliability, trust, and rapport, and how these perceptions in turn shape negative emotional responses. A 2×2 between-subjects experimental design manipulates tone and recovery strategy in chatbot-generated recovery messages. By observing user reactions to controlled AI failure scenarios, the study aims to clarify how communication congruence or mismatch affects cognitive and emotional responses. The findings will provide actionable insights for optimizing chatbot communication, contributing to a comprehensive understanding of AI-mediated service recovery and advancing theoretical and practical knowledge in this emerging area.
Comments
HCI