StoryBuddy: A Human-AI Collaborative Agent for Parent-Child Interactive Storytelling with Flexible Parent Involvement

Document Title

StoryBuddy: A Human-AI Collaborative Chatbot for Parent-Child Interactive Storytelling with Flexible Parental Involvement

Document Information

• Subject Area: Human-AI Collaboration, Conversational AI, Children's Educational Technology
• Keywords: Interactive Storytelling, Shared Reading, Dialogic Reading, Voice User Interface, Human-AI Collaboration, Child-Agent Interaction

Research Background and Issues

• Identified Problems or Challenges:

  1. Parents may face obstacles in interactive storytelling with their children due to limited time or difficulty in posing appropriate questions.
  2. Existing automated AI systems often do not support parental involvement, lack optimization for educational goals, and have low interactivity.
  3. Fully automated approaches to question generation and answering can reduce children's engagement and lack logical contextual question connections.

• Significance: Interactive storytelling has significant educational benefits for children's language fluency, communication skills, and cognitive development, while also enhancing parent-child relationships.

• Research Motivation and Related Work:

  1. Many digital tools have been used to support children's interactive storytelling but face limitations such as manual question design and lack of dynamic adaptation to user needs.
  2. With advancements in natural language understanding and question generation technologies, exploring human-AI collaboration can better meet dynamic needs and integrate parents' expertise.

Solution

• Proposed Method/Solution: The authors developed a system called "StoryBuddy" aimed at enhancing parent-child interactive storytelling experiences through human-AI collaboration.

• Innovations:

  1. Supports dynamic adaptation to user needs, allowing flexible switching between parent-assisted and AI-led storytelling modes.
  2. Automatically generates question-answer pairs for any story, while enabling parents to configure and review question content.
  3. Provides multi-level parental control and child development assessment features for different goals.

• Implementation Steps and Key Technologies:

  1. Shared Reading Mode: The system helps parents identify question opportunities, recommend appropriate questions, and create personalized follow-up questions.
  2. Automated Bot Reading Mode: Parents can predefine question generation model configurations, select the type of questions to generate, and let AI handle storytelling and interactive questioning.
  3. Progress Tracking and Assessment: The system displays children's performance data on a dashboard, helping parents evaluate their comprehension development.
  4. Technical Support:
     • Uses a BART model trained on the FairytaleQA dataset to generate question-answer pairs.
     • Applies the Google Cloud Dialogflow framework to handle task-oriented dialogues between children and AI.
     • Utilizes Google Speech API for seamless text-to-speech conversion.

Research Outcomes

• Specific Outcomes:

  1. Through human-AI collaborative design, the StoryBuddy system was successfully developed, optimizing the collaborative experience between parents and AI during storytelling.
  2. The system not only increased children's engagement but also supported flexible parental involvement and educational goal assessment.

• Advantages:

  1. Unlike traditional systems that can only handle pre-set stories, StoryBuddy generates interactive content suitable for any story.
  2. Offers greater parental flexibility, such as content configuration and mode selection, supporting both "synchronous" and "asynchronous" participation.

• Experimental or Evaluation Results: A user study involving 12 parent-child pairs revealed:

  1. Parents successfully used StoryBuddy to facilitate interactive storytelling.
  2. The system effectively balanced parental needs (e.g., educational and entertainment) with children's engagement across both modes.
  3. Users rated the system highly in terms of usability, practicality, and overall satisfaction.

• Limitations and Future Directions:

  1. Age Adaptability: Further optimization is needed to cater to the needs of children across different age groups, including adjusting question difficulty.
  2. Device Support: Adding support for screenless smart speakers to reduce screen time while addressing potential experience challenges with non-screen devices.
  3. User Base Expansion: Conducting larger-scale field tests to obtain more representative data.
  4. Feature Expansion: Supporting child-initiated questions and generating questions for image content to enhance interactivity and entertainment.

Conclusion

StoryBuddy provides an innovative human-AI collaborative approach to enhance interactive storytelling experiences by involving parents in configuration, collaborative storytelling, and assessment. Experiments demonstrate its effectiveness in optimizing parent-child interaction, supporting flexible parental involvement, and balancing educational and entertainment goals for children. The study also underscores the importance of defining appropriate AI roles in storytelling design.