A Narrative Agent for the “Family Story Hoard”: An Information-Theoretic Framework for Interactive Storytelling

Abstract

We introduce a narrative agent designed to facilitate the creation and preservation of personal stories within the "Family Story Hoard." The agent engages elderly users in interactive dialogues to elicit life stories, stores narrative elements, and evaluates the completeness of the story using information-theoretic metrics alongside Todorov’s five-stage narrative structure. The system dynamically guides users to fill narrative gaps through tailored prompts, ensures privacy by anonymizing sensitive data during storytelling, and automates archive with generated titles and chronological organization. By integrating conversational AI, information theory, and narrative analysis, this framework supports memory preservation while addressing challenges in coherence, completeness, and user privacy.