2026-10-14T11:59:48Z https://www.scidb.cn/oai

10.57760/sciencedb.psych.01058 2026-08-05T07:32:39Z

2026-08-05 Does Disclosing Negative Emotions to Generative AI Make People Feel Better? The Mechanisms and Boundary Conditions doi:10.57760/sciencedb.psych.01058 en Generative Artificial Intelligence (GenAI) are increasingly serving as a channel for individuals to disclose negative emotions. Although prior research has extensively examined AI applications in information delivery, task collaboration, and general social support, limited attention has been devoted to the emotional outcomes of disclosing negative emotions to GenAI. In particular, it remains unclear whether such disclosure effectively promotes emotional relief, through which psychological mechanisms these effects operate, and under what conditions they are strengthened. Drawing upon self-disclosure theory, and emotion regulation theory, the present research investigated the effectiveness, underlying mechanisms, and boundary conditions of negative emotional disclosure to GenAI.Three studies were conducted. In Study 1, a single-factor between-subjects experimental design was employed. After inducing negative emotions, participants were randomly assigned to one of three conditions: disclosure to GenAI, online expressive writing, or self-reflection. Participants engaged in a 10-minute interaction task followed by emotional state assessments. After attention checks, data from 160 valid participants were retained. Study 2 utilized a cross-sectional survey design, collecting 410 valid responses from individuals who had recently disclosed negative emotions to either GenAI or humans. Measures included emotional relief, perceived empathy, and cognitive reappraisal. Study 3 adopted a 2 (disclosure target: GenAI vs. human) × 2 (message contingency: high vs. low) between-subjects experimental design, following procedures similar to Study 1, with data from 134 valid participants retained. Studies 1 and 3 were implemented via the Credamo platform using the Doubao-seed-1-6-251015 large language model. All studies received approval from the institutional ethics committee, and informed consent was obtained from all participants. perceived empathy; cognitive reappraisal; message contingency; improvement yang xin xin Zheng Hang Zhang Chunmei Shi Yibo RESTRICTED dataset Science Data Bank