Accompany Sleep: Using GenAI to Create Bedtime Stories for Mediating Parent-Child Relationships in LBC Families

Research Background and Problem

• Issues and Challenges
  The authors focus on the psychological and behavioral issues of "Left-Behind Children (LBC)" in China. These children, lacking emotional support due to the prolonged absence of their parents, are prone to developmental delays and other psychological problems. Existing remote communication methods (e.g., video calls) fail to sufficiently foster effective emotional connections, primarily facing challenges such as limited topics, superficial interactions, and device usage constraints.

• Significance
  The issue of left-behind children affects tens of millions of families. Building deeper emotional and communicative bonds between parents and children is crucial for mitigating psychological harm to children and improving family relationships.

• Research Motivation and Related Work
  There have been attempts to improve interactions within separated families using technology (e.g., artificial intelligence), such as enhancing communication through haptic devices or electronic sandboxes. However, these methods are often limited to one-way information transmission or basic psychological assessments, failing to fully explore the bidirectionality and emotional depth of parent-child interactions.

Solution

• Proposed Method or Solution
  The authors developed a system called "Accompany Sleep," which transforms parents' daily life content into bedtime stories suitable for children using generative artificial intelligence (GenAI, such as ChatGPT4o) to create visual and auditory content. The system includes a "story creation application" for parents and a "projector device" for children.

• Innovations

  1. Personalized Generation: Stories are generated based on real-life fragments submitted by parents, tailored to children's actual interests and cognitive levels, enhancing the authenticity and emotional connection of the stories.
  2. Bidirectional Interaction: The system not only helps children better understand their parents' lives but also encourages children to actively interact with their parents.
  3. Enhanced Education and Emotional Support: Engaging stories convey important values, daily habits, and other content, while the projector's physical design supports children's emotional stability.

• Implementation Steps and Key Technologies

  1. Parents upload images and text content from their daily lives via a mobile application.
  2. The system utilizes a generative AI model (ChatGPT4o) to create complete stories, including text, accompanying images, and audio, based on the uploaded content.
  3. The stories are presented in a combination of visual and auditory formats through the children's projector device, fostering deep engagement and interaction.

Research Outcomes

• Specific Results

  • The system was tested with four left-behind families over one month, demonstrating enhanced family interactions and improved understanding and empathy between children and parents.
  • Data showed a significant increase in parent-child relationship scores (average score increased from 14.2 to 19.6, p-value = 0.0195).

• Advantages and Comparisons

  • Compared to existing communication tools, Accompany Sleep exhibited a deeper emotional connection capability, not only transmitting information through stories but also creating a unique interactive mechanism for parent-child relationships.
  • The automatically generated story content was both authentic and engaging, meeting children's cognitive needs while offering creativity and educational value.

• Experimental Results

  • The experiment revealed that parents actively curated content when uploading daily life fragments to enhance educational and positive aspects.
  • Children developed new bedtime habits, using the stories to initiate contact with their parents and express understanding and interest.
  • The system triggered positive behaviors, such as helping to correct children's bad habits.

• Limitations and Future Directions

  1. The sample size was small (only four families), limiting the ability to verify the system's adaptability to other cultural contexts and family models.
  2. The study relied on self-reported data, which may introduce subjective bias.
  3. There is room for improvement in the diversity of story generation formats and the feedback mechanism for children.
  4. Potential ethical issues with GenAI (e.g., generating false information) require further attention.

  Future Directions

  • Expand the sample size and conduct long-term follow-up studies to explore whether the system can achieve sustained improvements in parent-child relationships.
  • Introduce more objective evaluation methods (e.g., behavioral observation and interaction logs) to validate the results.
  • Enhance the emotional dimensions of the stories and explore the inclusion of parents' voice recordings.
  • Seek collaboration with governments or non-profit organizations to reduce costs and make the system accessible to more families.