"Did you sleep well?": A Multimodal Sleep Diary for Sustained Self-Reporting by Children

Research Background and Issues

• Identified Problems or Challenges:

  1. Sleep diaries are essential tools for assessing children's sleep patterns, but they face two significant challenges: maintaining long-term engagement and ensuring high-quality self-reports.
  2. Most existing digital sleep diary designs are tailored for adults, failing to adequately meet the needs of child users.
  3. Although voice input has shown potential in education and interaction domains, its practical effectiveness and long-term impact on children's conversational self-reporting remain unclear.

• Significance: The long-term effects of sleep disorders in children can significantly reduce their quality of life, impair learning abilities, and even correlate with other medical issues. Effective self-reporting tools can help doctors continuously monitor these problems and provide a basis for treatment.

• Research Motivation and Related Work:

  • Voice interaction, as a potential technology, can improve children's engagement and enhance report quality through a more user-friendly conversational interface.
  • Previous studies have explored the benefits of voice input for children's engagement but lack in-depth research on its long-term effects in real-world scenarios.

Proposed Solution

• Proposed Solution: The authors developed a multimodal sleep diary tool that integrates voice and text input, allowing children to choose their preferred input method.

• Innovations:

  1. Integration of voice and text input into a single multimodal application to enhance children's engagement and report quality in sleep diaries.
  2. Introduction of a linear mixed-effects model analysis based on actual usage to examine dynamic changes in modality preferences and reporting patterns.
  3. Systematic development of a Response Quality Index (RQI) that evaluates data quality through three metrics: information units, relevance, and clarity.

• Key Implementation Steps:

  1. Application Design and Development:
     • Constructed a multimodal conversational interface using a co-design approach with children.
     • Simplified and localized question sets to suit children's cognitive and language levels.
  2. Experimental Design:
     • Conducted a five-day field study collecting sleep diary data from 20 children aged 7-12, covering both text and voice input modes.
  3. Data Analysis:
     • Used linear mixed-effects models to analyze the quality and engagement of self-reports.
     • Employed quantitative metrics to evaluate children's input preferences and report quality.

Research Outcomes

• Specific Findings:

  1. Voice input was most beneficial for younger children, significantly increasing their engagement.
  2. Age was closely related to input preferences: younger children favored voice input, while older children preferred text input.
  3. Two distinct patterns of report quality changes were identified: a "U-shaped" pattern (decline followed by recovery) and a stable pattern.

• Advantages Compared to Existing Solutions:

  1. Clearly demonstrated the value of voice input in improving long-term engagement among younger children.
  2. Proposed a multimodal interaction design superior to traditional sleep diaries, better addressing children's diverse needs.
  3. Quantified five days of report content using an intuitive data quality evaluation system, revealing trends.

• Experimental or Evaluation Results:

  1. Responses generated via voice input had an average RQI higher than text input (approximately 0.54 units improvement).
  2. Although younger children preferred voice input, the quality of their results was still lower than older children's text input.
  3. Report length showed a gradual decline in engagement over five days, but voice input slightly recovered participation in later stages.

• Limitations and Future Directions:

  1. Limitations:
     • The sample size was small (only 20 participants), which may not fully represent a broader population of children.
     • The study duration was short (five days), failing to capture changes in long-term preferences and learning effects.
     • The potential impact of individual familiarity with technology on input preferences was not comprehensively studied.
  2. Future Directions:
     • Use larger samples and extend the observation period to several weeks or months to validate data stability.
     • Explore the potential of additional input modes (e.g., gesture or graphical input).
     • Investigate the adaptability of voice recognition technology for children from different linguistic or cultural backgrounds.

Conclusion

This study systematically explored the potential advantages and applicability of voice and text input in the context of children's sleep diaries for the first time. The work not only revealed the complex relationships between age, report quality, and modality preferences but also provided practical recommendations for child-friendly design. Future research should focus on broader populations, more complex tasks, and the long-term potential of multimodal systems.