A Framework for Emergent Narrative in Interactive Art Installations

Thesis Abstract

The study addresses the challenge of designing interactive art installations that autonomously generate coherent, user-responsive narratives without predefined scripts, enabling emergent storytelling that adapts to participant actions, physical space, and acoustic/visual context. Despite growing interest in interactive media, there remains a gap in formalized frameworks that integrate narrative theory, interaction design, and system architecture to produce verisimilar emergent narratives while preserving artistic intent and experiential accessibility. The aim is to develop a theoretical and practical framework—A Framework for Emergent Narrative in Interactive Art Installations—that unifies narrative emergence with real-time computation, perceptual cues, and collaborative audience engagement. Specific objectives are to (1) synthesize concepts from narratology, ambient intelligence, and human–computer interaction to define a generative narrative model for installations; (2) design and implement a modular, policy-based architecture that governs narrative progression through perception, action, and affective states; (3) empirically evaluate how different interaction modalities influence narrative coherence, engagement, and perceived agency; (4) validate the framework through iterative prototyping in three distinct gallery environments; and (5) articulate guidelines for artists and curators on balancing authorial intent with emergent systems. The methodology adopts a mixed-methods research design combining design-based research with formative and summative evaluation. The population comprises interactive installation artworks and their attending audiences across three urban museums in Europe and North America. A purposive sample of five completed installations and twelve new prototypes will be studied, drawing 60–80 participating visitors per prototype for quantitative measures and 20 in-depth case studies for qualitative insights. Data collection instruments include sensor-derived interaction logs (motion, gaze, proximity, device input), system state traces (narrative agent belief and goal trees, affective scoring from facial expression analysis and wearable galvanic skin response), post-experience questionnaires using a Likert scale for narrative satisfaction, semistructured interviews with artists and technical mentors, and workshop-based reflective notes. The instruments’ validity and reliability will be established through pilot testing with 15 participants and triangulation across data streams. Data analysis will employ a combination of thematic analysis for interview and open-ended responses, mixed-effects modeling to assess the impact of interaction modality on narrative engagement, and sequence analysis of narrative state transitions to evaluate coherence. A formal conceptual model will be specified using a goal–belief–action framework, operationalized through a policy-based narrative engine that orchestrates story arcs from perception inputs, agent deliberation, and multimodal output synthesis. Model validation will use abductive reasoning against observed emergent patterns, complemented by actor–network theory lenses to understand the distributed agency between human participants and computational narrators. Key expected findings include (a) a validated set of narrative policies and thresholds that reliably produce coherent micro-arc evolutions across diverse contexts, (b) empirical evidence that multimodal interaction (visuo-auditory input combined with haptic feedback) enhances perceived narrative agency and immersion, (c) identification of critical junctures where user actions can positively or negatively affect narrative continuity, and (d) a taxonomy of emergent narrative types (episodic, situational, and collaborative) with corresponding design principles. The study anticipates that narrative coherence will be highest when the engine adheres to constraining rules derived from narratology (agency, temporality, and goal-directedness) while allowing probabilistic variation in character behavior and plot events. Contributions to knowledge include a theoretically grounded framework that articulates how emergent narratives can be modularly designed, implemented, and evaluated within interactive art installations; a transferable architectural blueprint for artists and technologists that reconciles artistic autonomy with audience-driven variability; and an empirically informed typology of emergent narrative modes with practical guidelines for curatorial practice. The final recommendations will address design workflows, ethical considerations related to perceptual load and participant agency, and future directions for refining the policy engine with advances in affective computing and intelligent agents. The study concludes that a disciplined integration of narratology, human–computer interaction, and system architecture can yield robust emergent narratives that enhance engagement without sacrificing artistic intention, and it recommends iterative, cross-disciplinary collaboration and open-source tooling to foster broader adoption in contemporary art practice.