An Effective Multimodal Personality Detection Framework on Social Media

Personality detection is crucial for applications spanning mental health, marketing, and personalized education. Traditional detection methods rely on subjective self-reporting, suffer from inefficiency, subjectivity, and scalability constraints. The rise of social networks offers a transformative alternative, enabling implicit personality inference from user-generated content. However, current research predominantly focuses on textual analysis, overlooking interactions between multimodal information. Inspired by psychological theories, we extract finegrained image style features to capture visual manifestations of personality. We propose a novel multimodal framework for personality detection that integrates textual and visual modalities from social media platforms, which outperforms state-of-the-art unimodal baselines in rigorous experiments.