Cinematic Narratives as Socio-Technical Systems: Emotion Mining and Script-Audience Emotional Fidelity

Abstract

Cinema can be conceptualized as a socio-technical system in which scripts encode intended emotions, production processes transform them into multimodal experiences, and audiences generate emergent responses through reviews and ratings. This study investigates the emotional fidelity between designed affective trajectories in film scripts and perceived emotions expressed in audience reviews. A system-oriented computational framework was developed, integrating large-scale script and review data with transformer-based natural language processing models fine-tuned on the GoEmotions dataset. By applying a unified classification pipeline, we compare emotional distributions across scripts and reviews, analyze temporal and genre-specific patterns, and examine correlations with film success metrics such as profit and ratings. The results reveal both convergence and divergence between scripted intentions and audience responses, with genres functioning as semi-autonomous subsystems and historical trends reflecting context-dependent adaptation. Emotional fidelity-defined as the degree to which intended emotional expressions are preserved, transformed, or inverted in audience interpretation-is introduced as a system-level performance indicator. These findings advance theoretical perspectives on narrative communication as a feedback-driven socio-technical process and demonstrate how emotion mining can function as affective monitoring infrastructure for complex adaptive systems. The study contributes actionable insights for screenwriters, producers, and system designers seeking to enhance affective engagement.