Predictive Audience Engagement for Performing Arts

The primary indicator of performing arts impact and sustainability is the audience engagement, which is traditionally measured retrospectively and in coarse-grain, as surveys, attendance statistics, and critical reviews. The current paper suggests a human-oriented predictive audience engagement framework incorporating cognitive theory, multimodal behavioral and emotive predictors, and human-in-the-loop analytics into a real-time deployable system framework. Engagement is conceptualized as a multidimensional and temporal construct that is influenced by attentional, affective, interpretive and social processes. Temporal predictive models with uncertainty-sensitive inference are applied to multimodal audience data comprising of visual, acoustic, and contextual cues to predict engagement trajectories and estimate them over time. Key points in the process of human expertise are expert annotation, interpretive validation and ethical oversight, being transparent and having context validity. Experimental testing in a variety of performing arts contexts shows that predictive models based on time and multi-modal effects are superior to more basic predictive baselines in predicting engagement dynamics, especially in the context of prominent changes of performance. The expert evaluation of quality further ascertains the semantic congruence of the patterns of engagement predicted and artistic intent. The findings show that predictive analytics based on cognitive foundations and supplemented by human judgment can deliver useful and interpretable information to help in reflective practice, performance analysis, and audience-conscious artistic development

Analytics · Emotive · Performing arts · Predictive analytics · The arts · Thematic analysis · Creativity in Education and Neuroscience · Music Technology and Sound Studies · Neuroscience and Music Perception