This study examines consumer evaluations of robotic service failures caused by human interference by integrating service context, robot appearance, and individual anthropomorphism tendencies into a unified model. Two between-subjects experiments were conducted. In Study 1 (N = 402), participants interacted with a healthcare or food-service bot that failed due to verbal interference. Healthcare service failure elicited significantly more negative attitudes and lower failure tolerance than food service failure, and failure tolerance fully mediated the relationship between context and attitudes. In Study 2 (N = 213), we employed a 2 × 2 design (healthcare vs. food services × human-like vs. machine-like robot) and measured perceived deservingness and trait anthropomorphism. Human-like robots were judged most harshly when failing in healthcare (vs. food) services, whereas machine-like robots received similar evaluations across contexts. Perceived deservingness of the robot mediated this interaction. Moreover, the moderated-mediation effect occurred only among individuals with low to medium anthropomorphism tendencies. By positioning failure tolerance and deservingness judgments as core mechanisms in human–robot interaction, our findings advance theoretical understanding of moral attributions in service failure. Practically, they highlight the importance of matching robot anthropomorphic cues to service criticality: less human-like designs in high-stakes environments, while more human-like appearances may be appropriate in lower-stakes settings.

中文翻译：

---

从咖啡馆到诊所：消费者对类人和机器服务机器人故障的态度

本研究通过将服务环境、机器人外观和个人拟人化倾向整合到一个统一的模型中，考察了消费者对人为干扰导致的机器人服务故障的评价。进行了两个受试者间实验。在研究 1 （N = 402） 中，参与者与因语言干扰而失败的医疗保健或食品服务机器人互动。与餐饮服务失败相比，医疗保健服务失败引发的消极态度明显更多，失败容忍度更低，并且失败容忍度充分中介了环境与态度之间的关系。在研究 2 （N = 213） 中，我们采用了 2 × 2 设计（医疗保健与食品服务×类人机器人与类似机器的机器人）并测量了感知的应得性和特质拟人化。当医疗保健（与食品）服务失败时，类人机器人的评价最为严厉，而类似机器的机器人在各种情况下都受到了类似的评价。机器人的感知应得性介导了这种互动。此外，中等中介效应仅发生在具有中低拟人化倾向的个体中。通过将容错和应得判断定位为人机交互的核心机制，我们的研究结果促进了对服务失败中道德归因的理论理解。实际上，它们强调了将机器人拟人化线索与服务关键性相匹配的重要性：在高风险环境中，不太像人类的设计，而在低风险环境中，更像人类的外观可能更合适。