Proceedings of the Annual Meeting of the Cognitive Science Society

This study presents ADViRDS, an innovative scale and dataset specifically developed for examining the psychological traits of domestic violence (DV) perpetrators. Recognizing the critical need to understand the psychological dynamics of perpetrators, our research shifts the focus from the experiences of DV victims to the characteristics of the perpetrators. Our approach involves a six-dimensional scale designed to detect the psychological traits of DV perpetrators, formulated with insights from established DV research and psychologists. To complement this scale, we constructed a detailed dataset containing 574 individual entries from the Chinese social media platform "Zhihu." Each entry was carefully annotated by experienced professionals, ensuring a high degree of accuracy and relevance. We conducted a comprehensive analysis using a range of models, including Zero-Shot classification, GPT series, and fine-tuned pre-trained models, to evaluate their effectiveness in identifying individuals with psychological predispositions to DV. The findings reveal significant insights into the models' capabilities, highlighting the nuances in detecting DV tendencies through psychological profiling. Our research offers a new paradigm in DV studies, focusing on the psychological traits of perpetrators for a comprehensive understanding of DV dynamics and prevention.