Multimodal Emotion Recognition in Cinema: A "Synthetic" Human- and ChatGPT- Driven Analysis of Emotional Dynamics in Sophie's Choice.

This article examines the integration of artificial intelligence (AI), specifically ChatGPT, with human expertise to analyze emotional dynamics in cinema through a multimodal approach combining visual, auditory, and textual inputs. Focusing on a key scene from the film *Sophie's Choice*, two human judges segmented the scene into 56 sequences based on facial expressions, which then guided ChatGPT's simulated analyses of facial micro-expressions, dialogue sentiment, cinematographic elements, and sound cues using conceptual frameworks of established computational tools. The study found that ChatGPT could detect nuanced emotional states such as fear, despair, and authority, while acknowledging limitations in low-visibility conditions and the need for human supervision to ensure interpretive accuracy. This research highlights the potential of combining AI and human analysis to provide a richer, context-sensitive understanding of complex emotional narratives in film and suggests future directions including fully autonomous AI analyses and incorporation of viewer subjective experiences.