CharacterCritique: Supporting Children's Development of Critical Thinking through Multi-Agent Interaction in Story Reading

Research Background and Issues

• What problems or challenges did the authors identify?

  • Current education focuses more on children's understanding of factual knowledge and less on fostering critical thinking skills.
  • In family environments, parents often lack the skills to guide critical thinking when children read picture books, leading to superficial reading experiences.
  • Parents generally believe that fostering critical thinking is the responsibility of schools, neglecting the important role of the family.
  • Critical thinking education requires diverse methods to encourage higher-order thinking in children, but there is currently a lack of resources and methods to support parents.

• Why is this issue important?

  • Critical thinking helps children navigate a complex world and an uncertain future, forming the core of independent thinking habits and problem-solving abilities.
  • Early childhood is a critical period for developing critical thinking, and the formation of early thinking skills will influence long-term personal development.

• Research Motivation and Related Work

  • Literature shows that combining story reading with inquiry-based dialogue can effectively cultivate critical thinking.
  • Artificial intelligence and conversational agents are increasingly showing potential in the education field, but current research primarily focuses on adult users, with little attention to designs for fostering critical thinking in children.

Solution

• What methods or solutions did the authors propose?

  • Developed a multi-agent interactive system called CharacterCritique, which promotes critical thinking in children within family environments through story reading and dialogue with virtual characters.
  • The system integrates multi-agent characters supported by a large language model (GPT-4o) to help children analyze problems, explain, and evaluate viewpoints through character "conversations."
  • Replaces traditional binary Q&A with open-ended questions and character dialogues to stimulate deeper exploration by children.

• What are the innovative aspects of this solution?

  • Introduces multiple characters (e.g., a fisher village girl, a rural boy), each representing different perspectives, to provide feedback and challenges to children's solutions.
  • Supports children in creating custom agents (e.g., an engineer or a mayor) to interact with specific characters, acquire knowledge, and expand cognition.
  • Dynamically generates story endings based on children's final decisions, adding fun and feedback to the learning process.

• What are the implementation steps and key technologies used?

  1. Story Design: Selected challenging stories suitable for 5-6-year-old children, based on complex themes like ethics, distribution, and cooperation.
  2. Multi-Agent Design: Characters were illustrated as cartoons using Midjourney, with dialogues generated by GPT-4o and customized prompts tailored to character traits.
  3. Human-Computer Interaction Design:
     • Designed open-ended Q&A mechanisms to encourage in-depth exploration.
     • Enabled children to choose interactions with different agents after reading, gradually refining their solutions.
  4. Experimental Design: Conducted a controlled experiment comparing traditional story reading to evaluate the system's impact on children's critical thinking through quantitative and qualitative analysis.

Research Outcomes

• What specific outcomes were achieved?

  • In a comparative experiment involving 32 parent-child pairs, the system significantly fostered children's critical thinking, particularly in the creativity and diversity of problem-solving.
  • Children's solutions became more comprehensive and detailed after interacting with virtual characters.
  • The system encouraged deeper parent-child dialogues, shifting parents' roles from "leaders" to "observers" and "supporters."

• What advantages does it have compared to existing solutions?

  • CharacterCritique provides open-ended and multi-character interaction, which stimulates deeper thinking in children compared to traditional picture story reading.
  • The multi-agent design introduces diverse perspectives, enabling children to refine their viewpoints through critical feedback.
  • Supports personalized learning experiences, such as generating background knowledge or diverse story endings through agents.

• What were the experimental or evaluation results?

  • Quantitative Analysis:
    • Children in the experimental group demonstrated higher critical thinking skills, such as analysis, reflection, and reasoning.
    • The experimental group used an average of 5.25 skills in problem-solving (compared to 2.94 in the control group).
  • Qualitative Feedback:
    • Both parents and children highly appreciated the interaction experience with agent characters, particularly the system's real-time feedback and flexible discussion format.
    • The system significantly increased the diversity of parent-child discussion topics (experimental group averaged 8.06 topics, compared to 3.25 in the control group).

• Limitations and Future Directions

  • Limitations:
    1. The system relies on static images and text, lacking dynamic content for a more immersive experience.
    2. Absence of long-term observational data, leaving the system's sustained impact on children's thinking development unclear.
    3. Current speech recognition technology is limited, with children's ambiguous oral expressions reducing system accuracy.
  • Future Directions:
    1. Improve visual design and introduce virtual animated characters to enhance immersion.
    2. Expand the story library and support content personalization, adjusting story difficulty and agent question complexity based on children's age.
    3. Enhance multimodal interaction capabilities, enabling agents to recognize children's non-verbal expressions (e.g., gestures).

Conclusion

The CharacterCritique system effectively cultivates children's critical thinking through multi-agent dialogue and parent-child interaction in family settings. Its diverse character interactions and open-ended question design not only enhance children's creativity and depth of thought but also strengthen communication and collaboration between parents and children. Future research should address the lack of immersion and unclear long-term effects while optimizing technical performance to further advance the system's educational applications.