Dark Romance as an AI-Mediated Experience: Identification with Characters and Attitudes towards Romantic AI

Abstract

The identification with fictional characters has been widely recognized as a powerful psychological phenomenon, influencing self-perception and attitudes toward interpersonal relationships. Within this context, the dark romance genre, characterized by emotionally charged narratives involving ambivalence, power imbalances, and problematic relational behaviors, provides a unique framework for examining how engagement with such narratives impacts attitudes toward intimacy and relational norms. Concurrently, the rising prevalence of AI-mediated conversational agents introduces new dimensions to the study of emotional and romantic connections, raising critical questions about human-AI intimacy and its implications. This study sought to explore the associations between identification with dark romance characters, emotional and sexual experiences involving AI systems, and attitudes toward problematic relational behaviors. A sample of 139 Romanian participants was recruited to investigate these dynamics. Participants completed the Character Identification Scale, the Experiences in Human–AI Relationships Scale (EHARS), selected subscales of the Intimate Partner Violence Attitude Scale–Revised (IPVAS-R), and newly developed measures assessing AI sexual satisfaction, emotional support provided by AI, positive attitudes toward romantic AI entities, and recognition of relational red flags. Mediation analyses indicated that identification with dark romance characters was indirectly associated with greater AI sexual satisfaction through more positive attitudes toward romantic AI, supporting a full mediation model. In contrast, identification did not mediate the relationship between AI emotional support and AI sexual satisfaction, which was characterized by a strong direct association.Moderation analyses revealed that recognition of relational red flags significantly attenuated the positive association between AI sexual satisfaction and victims' acceptance of abusive behaviors, functioning as a protective cognitive factor. No significant association was observed between AI sexual satisfaction and victims' acceptance of physical violence. Overall, the findings suggest that narrative identification and AI-related intimacy are linked through attitudinal mechanisms, while awareness of relational red flags plays a crucial role in limiting the normalization of abusive relational dynamics. These results highlight the importance of examining both media-based identification processes and cognitive protective factors in understanding emerging forms of human–AI intimacy.