"I Need Your Help!" : Facilitating Psychological Communication Between Left-Behind Children and Their Parents with an AI-Powered Sandbox

Research Background and Issues

• Issues or challenges identified by the authors:

  • In impoverished areas, limited resources, economic burdens, and insufficient psychological education for guardians lead to inadequate mental health support for children.
  • The prolonged absence of migrant worker parents exacerbates mental health problems, with parents generally lacking understanding of their children's psychological states.
  • Existing AI tools for children's mental health often overlook the specific needs of left-behind children and lack direct parental involvement.

• Why this issue is important:

  • Left-behind children, due to the absence of family companionship and psychological support, are prone to feelings of loneliness, anxiety, and depression, negatively impacting their growth and social adaptability.
  • The imbalance of medical and psychological health resources between urban and rural areas makes traditional psychological services prohibitively expensive, further limiting mental health interventions.

• Research motivation and related work:

  • Mental health issues among left-behind children have become a significant societal concern, calling for low-cost, scalable intervention methods.
  • Sandplay therapy, as a non-verbal psychological intervention method, is suitable for expressing children's inner emotions. However, traditional sandplay therapy involves high costs and professional guidance, making it difficult to implement in resource-limited areas.
  • The authors aim to leverage AI technology and family involvement to fill the gap in mental health services, providing psychological support for left-behind children and their remote guardians.

Solution

• Proposed solution by the authors:

  • Developed a low-cost digital sandplay tool (DiSandbox) integrating artificial intelligence for children's mental health assessment.
  • The system includes a physical sand tray, an AI assistant, and a mobile application that allows parents to view their children's sandplay creations and psychological state analyses.

• Innovative aspects:

  • Utilizes large language models to analyze children's sandplay creations, providing psychological state assessments and reducing reliance on professional psychologists.
  • Offers real-time viewing and suggestions for remote-working parents, enabling them to participate in their children's mental health interventions.
  • Optimizes the sandplay tool to suit economically disadvantaged families, achieving low cost and easy accessibility.

• Implementation steps and key technologies:

  1. Physical sandplay tool design: Uses portable, low-cost wooden sand trays and basic sandplay toys, along with blank toys to encourage children's creative expression.
  2. AI assistant: Records the creation process via voice and camera, employing YOLOv8 for toy classification and EmoLLM for psychological state analysis.
  3. Parental platform: A smartphone application displays sandplay psychological analysis reports and provides communication suggestions, such as open-ended questions and expressions of care.
  4. Sticker printing and interaction: Supports instant generation of sandplay elements and printing for children's use, enhancing creative interest.
  5. Experimental process: Designed a five-week experiment combining children's psychological tests (CDI, SCARED), sandplay analysis, and semi-structured interviews to validate the tool's effectiveness.

Research Outcomes

• Specific outcomes:

  • Experimental results show that DiSandbox significantly reduces depression and anxiety scores among left-behind children (CDI and SCARED tests).
  • Interactive and guided features improve children's self-expression and emotional regulation abilities.
  • The system successfully provides remote parents with psychological state reports and communication examples, helping them better understand and support their children.

• Advantages compared to existing solutions:

  • Costs approximately $40, significantly reducing expenses compared to traditional sandplay therapy.
  • AI-driven psychological assessments and suggestions reduce dependence on professional psychologists, enabling left-behind children to receive mental health support in family and school environments.
  • Enhances the content and quality of psychological communication between parents and children, enriching family interaction patterns.

• Experimental or evaluation results:

  • Experimental data indicates comprehensive improvement in psychological health indicators for participating children through sandplay activities.
  • High alignment with expert psychological assessments, with system accuracy scoring 0.79.
  • Both children and parents expressed satisfaction with system feedback and communication suggestions, with family communication frequency significantly increasing.

• Limitations and future directions:

  • Current sandplay psychological state assessment models are biased toward Chinese cultural contexts; future work should expand multicultural data to improve analysis effectiveness.
  • AI psychological interpretation still faces individual and situational differences, requiring further validation from psychological perspectives.
  • User interface optimization is needed to lower technological barriers and enhance parental involvement.
  • Incorporate long-term data tracking and health management features to provide more comprehensive psychological support solutions.

Conclusion

This study introduces the DiSandbox system as a low-cost, scalable mental health support tool for left-behind children and their families, demonstrating significant effectiveness in improving children's mental health and fostering parent-child communication. Future work should focus on enhancing cross-cultural applicability, optimizing user experience, and promoting collaboration with local education and psychological service institutions to expand the tool's societal impact.