From Static Images to Dynamic Interactions: Generative AI Exhibits Robust Social-Cognitive Capabilities across Modalities and Cultural Boundaries

We systematically evaluated the social-cognitive capabilities of state-of-the-art Generative AI models using a suite of textual, static image, and dynamic video assessments. Across all modalities, advanced models consistently met or exceeded average human performance, demonstrating robust, cross-ethnic abilities, suggesting their potential as objective clinical assessment tools.