Modeling Synthetic Audience Reactions to Social-Cognitive Narratives: A Generative GSR Model to Predict Real-World Autonomic Alignment during Film Viewing

Abstract

While media consumption can be a solitary act, it produces a shared, socially coordinated experience where audiences bodies align in response to shared narrative events that are often social-affective in nature. Despite this recognition, traditional descriptive models of Galvanic skin response (GSR) have existed for decades, yet the socially coordinated aspect remains to be fully reflected in physiological models with the field of communication often treating the underlying generators of autonomic activity as a black box. To bridge this gap, we introduce a computational framework that models the underlying neural driver and its convolution to sweat gland physiology to explain how narrative events translate into measurable conductance. By leveraging multimodal AI models to "interpret" the social-cognitive content of a film, we generated a predictor timeline for a synthetic audience comprised of digital agents (i.e. artificial body systems responding to the film events with GSR responses). We then test this computational audience model by comparing its predictions against an empirical dataset collected as audience members (N = 96) processed the same stimulus, finding that AI-identified social triggers, like moments of comedic violence or shared emotional shifts, significantly predict the GSR time-course of audience engagement. In sum, this paper moves beyond simple and often retrospective labels like "arousal" to offer a computational account of how shared social narratives grip the human nervous system. We provide a scalable and expandable framework and a set of tools to predict media impact and understanding the psychophysiological basis of media.