Characterizing LLM-Empowered Personalized Story Reading and Interaction for Children: Insights From Multi-Stakeholders' Perspective

Research Background and Issues

• What problems or challenges did the authors identify?
  Current AI-driven children's story reading tools still have limitations in supporting personalized interaction and dialogue. Existing tools mainly rely on pre-generated questions or fixed structured dialogue patterns, which fail to fully address children's dynamic needs, such as personalized interactions related to their interests, cognitive abilities, and family environments. Additionally, parents expect these tools to pose expansive questions that foster children's critical thinking and creativity.

• Why is this issue important?
  Personalized interaction is an indispensable part of children's story reading, as it enhances language and social skills while contributing to cultural and emotional awareness. Moreover, despite the widespread application of AI technologies, there is still a lack of comprehensive exploration into how large language models (LLMs) can effectively support children's personalized interaction needs. Addressing this issue not only holds academic significance but also contributes to the advancement of educational technology.

• Research Motivation and Related Work
  The research motivation stems from addressing the knowledge gap regarding how to effectively and appropriately use LLMs to enhance children's story reading experiences. By reviewing literature in related fields, the authors found that earlier efforts focused on optimizing technical performance or designing interactive tools, with few systematic studies on the need for personalized interaction support in children's story reading scenarios. The study draws on theories and practices related to children's education, AI, and HCI.

Solution

• What methods or solutions did the authors propose?
  The authors designed and developed a personalized interactive story reading tool called StoryMate, powered by large language models (LLMs). StoryMate aims to achieve child adaptation, dynamic adjustment, and personalized interaction through customized book content and reading modes, personalized chat, attention-capturing mechanisms, and guided dialogue based on RAG (retrieval-augmented generation).

• What are the innovative aspects of this solution?

  1. Customization Features: Supports children and parents in uploading and categorizing books, as well as adjusting interaction modes (e.g., setting interaction frequency).
  2. Real-time Adaptive Interaction: Utilizes GPT-4 to generate personalized questions and dialogue content based on children's age, interests, and real-time reactions.
  3. Guided Expansive Dialogue: Integrates external knowledge with story content through RAG technology, further stimulating children's active thinking and expanding their knowledge.
  4. Attention-Capturing Mechanism: Incorporates children's names, topics of interest, and encouraging feedback, combined with multimodal elements in a graphical user interface (GUI) to deeply engage children's interest in dialogue.

• What are the implementation steps and key technologies used?

  1. Tool Design: Front-end developed using HTML, JavaScript, and CSS, with Flask managing the back-end.
  2. Knowledge Integration: Knowledge base constructed based on NGSS (Next Generation Science Standards), using a fine-tuned Retriever model for knowledge matching and GPT-4 for generating guided dialogues.
  3. UI Optimization: Designed a child-safe interface (e.g., soft colors and simple layouts) supporting multimodal interactions such as text, images, and sound.

Research Outcomes

• What specific outcomes were achieved?

  1. The design and application of StoryMate provided rich personalized story reading and interactive features, receiving positive feedback from children, parents, and education experts.
  2. The tool supports expansive dialogues that inspire children's active thinking and deeply integrate their interests with learning content.
  3. Experiments showed that most children were able to engage with the interactive content generated by StoryMate and were enthusiastic about using the tool as a reading companion.

• What advantages does it have compared to existing solutions?

  • Dynamic Adaptability: Unlike existing tools that primarily rely on fixed templates for interaction, StoryMate demonstrates excellent flexibility and adaptability, adjusting in real-time based on user feedback.
  • Intelligent Expansion: Embeds external knowledge through RAG technology, creating knowledge-rich dialogues that go beyond the story narrative.
  • Emotional Engagement: Supports personalized feedback and naming mechanisms, helping establish emotional connections between children and the tool.

• What were the experimental or evaluation results?
  User experiments showed that most children could interact smoothly and exhibited high levels of interest in the tool. Parents and education experts noted that the tool effectively stimulated children's interest in exploring knowledge, particularly in questioning and guidance. However, feedback also indicated that some dialogue content might not fully match the cognitive abilities of all children.

• Limitations and Future Directions

  1. Some children reported that certain vocabulary or questions were too complex, especially younger participants who found interactions challenging.
  2. Concerns were raised about whether over-reliance on reward mechanisms might diminish children's intrinsic interest in reading.
  3. Current experimental participants were mainly from economically developed regions; future studies should expand to diverse social backgrounds to enhance research generalizability.
  4. Future efforts should optimize the tool's multimodal design and increase child-led interaction modes to improve fairness and autonomy.

Through the study of StoryMate, this paper provides design suggestions and research directions for future LLM-based interactive technologies aimed at children. These recommendations can help developers better address children's needs, reduce ethical risks in technology applications, and better integrate technology with educational practices.