Ask, Answer, and Detect: Role-Playing LLMs for Personality Detection with Question-Conditioned Mixture-of-Experts
ROME is a framework that improves personality detection by injecting psychological knowledge using LLMs' role-playing ability. It simulates users answering standardized questionnaires based on their social media posts, generating evidence that links language to specific traits. This approach outperforms existing methods by providing interpretable, questionnaire-grounded supervision.