Master's thesis / Computer Science
Active Inference for Affective and Narrative Interaction: SEPHIROT® as a Computational Framework for Human-AI Co-Creation
The thesis investigates narrative continuation as decision-making under uncertainty. Using SEPHIROT® and The Fox and the Crow as a bounded case study, it examines how Active Inference can support participant agency while respecting author-defined narrative constraints.
The work develops an inspectable computational approach to narrative control, supported by diagnostics and recorded participant interactions. It also identifies open questions in learning dramaturgical structures and evaluating their contribution to human–AI co-creation.
Read the abstract
Interactive narrative systems face the challenge of reconciling participant agency with authorial intention, particularly when human interventions depart from anticipated story paths. This thesis investigates whether Active Inference can provide a computational foundation for adaptive narrative control by framing narrative continuation as decision-making under uncertainty.
The research explores the relationship between computational narratology, generative models, and Expected Free Energy minimization, examining how learned event predictions and inferred latent states can contribute to the evaluation of narrative continuations within author-defined constraints.
A computational architecture is developed and examined through SEPHIROT®, a framework for participatory storytelling, using The Fox and the Crow as a bounded case study. The implementation integrates probabilistic inference, narrative-state representation, and validated continuation selection, supported by computational diagnostics and recorded participant interactions.
The thesis establishes an inspectable approach to inferential narrative control while identifying the conceptual and empirical challenges that remain in learning dramaturgical structures and evaluating their contribution to human–AI co-creation.
Keywords: Active Inference; computational narratology; interactive narrative; narrative control; Expected Free Energy; human–AI co-creation.
Citation: Anglani, Alessandro. Active Inference for Affective and Narrative Interaction: SEPHIROT® as a Computational Framework for Human-AI Co-Creation. Computer Science Master's Degree, academic year 2025–2026.
Downloads
Loading statistics…
