How do failure attribution and linguistic style affect the remedial effects of generative AI services? The importance of humorous self-deprecation – Evidence from ERPs
Dong Lv, Rui Sun, Jiajia Zuo, Qiuhua Zhu, Shukun Qin · Behaviour and Information Technology · 2025
In the highly competitive market for generative Artificial Intelligence (AI) products, effective remediation after the failure of AI product services is especially important. The aim of this study is to investigate the impact of attributions for failures in Generative AI services (whether attributed to the user, AI, or the environment) and different language styles (rational explanation or humorous self-deprecation) on users’ willingness to accept these services, including the interaction between these factors and the underlying automated cognitive processing mechanisms. Based on Dual-Processing Theory and Benign Violation Theory, explores changes in user attitudes toward the effectiveness of remediation for generative AI service failures through an electroencephalogram (EEG) experiment. Behavioral results found that generative AI explanations attributing failures to environmental factors rather than to the user or AI service had better remediation effects, and the use of a humorous and self-deprecatory language style can greatly improve remediation effectiveness. EEG results revealed that, compared to rational explanations, the humorous, self-deprecatory style elicited a larger N400 component, whereas rational explanatory style elicited a larger late positive potential (LPP) component. Under the humorous style, the N400 component induced by failure attribution to AI was significantly higher than that attributed to environmental factors or users.