This study investigates the efficacy of generative artificial intelligence in online service recovery; specifically, the use of ChatGPT (vs. human employees) in preparing managerial response(s) (MR or MRs) to online hotel reviews is considered. ChatGPT could be used to generate human-like MRs for online service recovery but this could backfire due to algorithm aversion when an individual discounts algorithm decisions relative to human-made decisions. Data collected via interviews, a modified Turing test and an online experiment provide empirical support for this. Findings reveal that potential customers could not clearly differentiate between the two types of MR and could not clearly identify the ‘better’ of the two. Yet, when informed of the MR source, ChatGPT MRs led to lower affective, cognitive and conative outcomes. Findings also unveiled perceived authenticity and uncanniness as significant parallel mediating pathways in this algorithm aversion. Theoretical and managerial implications are discussed.

中文翻译：

---

ChatGPT 和在线服务恢复：潜在客户对管理层对负面评论的反应


本研究调查了生成式人工智能在在线服务恢复中的功效;具体来说，考虑了使用 ChatGPT（与人类员工相比）来准备对在线酒店评论的管理响应（MR 或 MR）。ChatGPT 可用于生成类似人类的 MR 以进行在线服务恢复，但当个人相对于人为决策不重视算法决策时，这可能会适得其反，因为算法厌恶。通过访谈、改进的图灵测试和在线实验收集的数据为此提供了实证支持。研究结果显示，潜在客户无法清楚地区分两种类型的 MR，也无法清楚地识别两者中 “更好 ”的。然而，当被告知 MR 来源时，ChatGPT MR 导致情感、认知和共源结果降低。研究结果还揭示了感知的真实性和不可思议性是这种算法厌恶中重要的平行中介途径。讨论了理论和管理意义。