作者
Zijun Yan,Jiaqi Sun,Fangyu Song,Qingju Deng,Tong Chen,Jiatong Pan,Qiang Cao,Liangjing Xia
摘要
Background Transgender women in China face high rates of depression due to minority stress, family rejection, and limited access to gender-affirming care. AI chat tools are increasingly used for emotional support, yet their role in AI-mediated psychological self-rescue remains underexplored.Methods We employed a qualitatively driven mixed-methods design, conducting semi-structured interviews with 39 transgender women in China experiencing severe depression and using AI chat tools for emotional support. The analysis integrated human-led qualitative coding with quantitative indicators derived from recurrent experiences, applying conservative overlap analysis for benefit-harm coexistence and a bounded GPT-4o API-assisted memoing step after human coding had been completed.Results AI use was frequent, with 24/39 participants (61.5%) using it daily, mostly late at night. Four key functions emerged: relational holding, identity mirroring, crisis scaffolding, and self-governance. Benefits included reduced loneliness (33/39, 84.6%), learning coping strategies (29/39, 74.4%), and support during suicidal crises (28/39, 71.8%). However, challenges like privacy and data-security concerns (27/39, 69.2%) and misgendering (22/39, 56.4%) persisted. A conservative overlap analysis revealed that benefits and harms coexisted in a significant portion of the sample.Conclusion AI chat tools served as critical psychological lifelines for transgender women, offering emotional support, recognition, and crisis management in the absence of reliable human care. While valuable, they also presented risks of over-reliance, privacy concerns, and algorithmic harm. The study emphasizes the need for trans-affirming design, localized crisis support, and integration with community-based human resources.