Interactive Children’s Story Rewriting Through Parent-Children Interaction

Yoon-Joo Lee, Tae Soo Kim, Minsuk Chang, Juho Kim2022 · DOI

Abstract

Storytelling in early childhood provides significant benefits in language and literacy development, relationship building, and entertainment. To maximize these benefits, it is important to empower children with more agency. Interactive story rewriting through parent-children interaction can boost children's agency and help build the relationship between parent and child as they collaboratively create changes to an original story. However, for children with limited proficiency in reading and writing, parents must carry out multiple tasks to guide the rewriting process, which can incur high cognitive load. In this work, we introduce an interface design that aims to support children and parents to rewrite stories together with the help of AI techniques. We describe three design goals determined by a review of prior literature in interactive storytelling and existing educational activities. We also propose a preliminary prompt-based pipeline that uses GPT-3 to realize the design goals and enable the interface.

Introduction

Storytelling is a cornerstone of early childhood development. "Story rewriting"—where a child modifies elements of an existing narrative—is particularly powerful for building agency and literacy. However, children in pre-reading or early-reading stages (ages 3–8) cannot perform this task alone. When parents facilitate this process, they face a high cognitive load, needing to identify changeable story parts, generate engaging questions, and rewrite the narrative on the fly while maintaining grammatical and logical consistency.

This paper introduces an AI-assisted interface designed to help parents and young children collaboratively rewrite stories. By leveraging GPT-3 to handle complex narrative updates and generate conversational prompts, the system aims to boost a child’s sense of agency while reducing the parent's cognitive "multitasking" load.

Methods and Pipeline

The system utilizes a three-step pipeline powered by GPT-3 and few-shot prompting to facilitate the rewriting process.

  1. Span Extraction: The system identifies "changeable" entities within the story across six dimensions: Character, Setting (Time/Place), Character Description, Feeling/Emotion, and Action.
  2. Question Generation: For a selected span (e.g., a character's name), the AI generates several questions the parent can ask the child (e.g., "Who is your best friend in kindergarten?"). This scaffolds the parent-child dialogue.
  3. Story Rewriting: Once the child provides an answer, the AI rewrites the paragraph. This includes Coreference Resolution (updating pronouns) and maintaining Semantic Consistency (e.g., changing "Gumbo" to "Bibimbap" if the setting changes from New Orleans to Seoul).
The GPT-3 based pipeline for span finding, question generation, and story rewriting.

User Interface and Interaction

The interface is designed to treat the parent as the primary user who "filters" AI suggestions for the child. The layout includes an "Original Story" panel, a "Current Span" panel for question selection and answer input, and a "New Story" panel where the rewritten text appears.

The interactive interface showing the original story, question generation, and the rewritten story with human-in-the-loop acceptance.

A key feature is the Human-in-the-Loop mechanism. When the AI suggests secondary changes to maintain semantic consistency (like the "Gumbo" example), the interface allows parents to "Accept" or "Deny" these changes. This ensures parents maintain control over the narrative flow and quality.

Key Insights and Evaluation Plan

The design focuses on three psychological pillars of agency: autonomy (freedom to choose), competence (feeling capable), and effectance (seeing the impact of choices). By automatically suggesting related changes, the AI helps the child see the "global effectance" of their creative decisions.

While quantitative results are pending, the authors propose a robust evaluation plan:

  • Technical Evaluation: Benchmarking entity extraction against the FAIRYTALE QA dataset.
  • Human Evaluation: Expert scoring of generated questions for readability and relevance.
  • User Study: Observing parent-child pairs to measure the enhancement of the child's sense of agency and system usability.

Conclusion and Future Work

The system aims to balance automation with human control. Future work will address model biases in LLMs and explore multi-turn dialogue to create a more natural storytelling flow. Additionally, the researchers plan to refine the UI to ensure it remains intuitive for parents during active interaction with their children.

Summary updated

Abstract

Storytelling in early childhood provides significant benefits in language and literacy development, relationship building, and entertainment. To maximize these benefits, it is important to empower children with more agency. Interactive story rewriting through parent-children interaction can boost children's agency and help build the relationship between parent and child as they collaboratively create changes to an original story. However, for children with limited proficiency in reading and writing, parents must carry out multiple tasks to guide the rewriting process, which can incur high cognitive load. In this work, we introduce an interface design that aims to support children and parents to rewrite stories together with the help of AI techniques. We describe three design goals determined by a review of prior literature in interactive storytelling and existing educational activities. We also propose a preliminary prompt-based pipeline that uses GPT-3 to realize the design goals and enable the interface.
BibTeX
@misc{lee2022interactive,
  author = {Yoon-Joo Lee and Tae Soo Kim and Minsuk Chang and Juho Kim},
  title = {{Interactive Children’s Story Rewriting Through Parent-Children Interaction}},
  year = {2022},
  doi = {10.18653/v1/2022.in2writing-1.9},
}
Figure