Interactive AI-driven dramaturgy for adaptive live theatre experiences

Thesis Abstract

This study addresses the increasingly dynamic interface between live theatre and intelligent dramaturgy by examining how AI-driven adaptive dramaturgy systems can modify performance material in real time to respond to audience cues while preserving artistic coherence. The problem centers on balancing automated adaptability with narrative integrity, ensuring ethical deployment, and evaluating audience reception across variable performance paths. The aim is to design, implement, and evaluate an integrated AI dramaturgy framework that modulates scene pacing, character behavior, and lighting cues in response to live audience signals without compromising dramaturgical goals. Specific objectives include (i) to develop an adaptive dramaturgy architecture that integrates natural language processing, computer vision, and recommender-based decision logic; (ii) to operationalize a set of theatrical constraints that preserve character voice, thematic arc, and stage aesthetics; (iii) to assess the impact of AI-driven adaptations on audience engagement, perceived authenticity, and memory retention; (iv) to evaluate performer workload and workflow implications during live performances; and (v) to propose a governance model addressing ethical and artistic accountability. The methodology adopts a mixed-methods design within a design-based research framework. The study will be conducted in three stages across two professional theatre productions in a metropolitan venue with a combined audience of approximately 1,800 attendees. Stage 1 involves the development of a modular AI dramaturgy system comprising a scene planner, a dialogue variation engine, and an adaptive stage management module, guided by theories of ludic dramaturgy and adaptive narrative (genealogically rooted in Mieke Bal’s narrative theories and Janet Murray’s “Plausible World”). Stage 2 entails a controlled field trial across 12 performances, with 300 audience participants and 32 professional actors and technicians as participants in the evaluation. Data collection instruments include (a) biometric and gaze-tracking sensors for audience arousal and attention, (b) post-show questionnaires using a Likert-scale and open-ended items to gauge engagement and perceived coherence, (c) expert dramaturgical assessments using a validated rubric, and (d) semi-structured interviews with performers and technical staff. Validity and reliability will be established through triangulation, pilot testing of the AI modules (n=3 workshops), and inter-rater reliability checks on dramaturgical assessments (Cohen’s kappa > 0.75). Data analysis will employ a combination of statistical and qualitative techniques. Descriptive statistics will summarize audience engagement metrics; repeated-measures ANOVA will examine differences in engagement and recall across adaptive versus non-adaptive scenes; regression analysis will explore predictors of perceived authenticity and satisfaction. Thematic analysis will be applied to interview data to identify tensions between automated adaptation and artistic intent, with coding guided by the theoretical framework. A qualitative–quantitative synthesis will be conducted using a convergent mixed-methods approach to triangulate findings. The study will also include a model specification for the adaptive system’s decision logic, including constraint satisfaction formulations to maintain dramaturgical coherence under varying audience states. Expected findings include (i) measurable increases in audience engagement and recall in adaptive scenes compared to fixed-script benchmarks; (ii) identification of threshold conditions under which AI interventions enhance or hinder perceived authenticity; (iii) evidence of workable task distribution between actors and the AI dramaturgical agent that preserves performance flow; and (iv) practical guidelines for ethical governance, accountability, and artistic control in AI-assisted theatre. The contribution to knowledge encompasses establishing a rigorous, field-tested framework for AI-driven dramaturgy that foregrounds artistic integrity and performer agency; advancing empirical understanding of audience adaptability to AI-modulated performances; and providing a scalable blueprint for theatres seeking to implement adaptive storytelling technologies. The study concludes that well-designed AI dramaturgy can augment live theatre by offering personalized yet coherent audience-centric experiences, provided that dramaturgical constraints are explicit, performer training is integrated into workflow, and robust ethical governance is maintained. Recommendations include developing standardized performance-ethics guidelines, investing in cross-disciplinary training for creative teams, and exploring long-term implications for repertoire programming and audience development.