Self-E: Smartphone-Supported Guidance for Customizable Self-Experimentation

Document Title

Self-E: Smartphone-Supported Guidance for Customizable Self-Experimentation

Document Information

• Subject Area: Smartphone-based self-experimentation and personalized health behavior analysis
• Keywords: self-experimentation, personal informatics, self-tracking, lifestyle improvement, user guidance, data visualization, customizable experiments, user trust, mobile health, experimental design

Research Background and Issues

• Identified Problems/Challenges:
  • While self-tracking devices and related applications support health improvement and behavior adjustment, most ordinary users lack the ability to analyze personal data, making it difficult to design and execute controlled lifecycle experiments.
  • Existing self-experimentation methods and systems either lack guidance on causal relationships or are limited to specific domains or populations.
• Significance:
  • Personalized health strategies need to uncover potential causal relationships based on individual lifestyles and behavioral characteristics, but traditional group randomized controlled trials (RCTs) are unsuitable for scenarios with significant individual differences.
• Research Motivation and Related Work:
  • The authors aim to develop a system that lowers the barrier to self-experimentation, enabling users to conduct cross-domain experiments more conveniently in their daily lives.
  • Design goals include high user freedom, experimental guidance, and reducing the learning curve, allowing users to quickly get started and obtain meaningful results.

Solution

• Proposed Solution:
  • The authors developed a mobile application called "Self-E" that guides users through self-experimentation, including experimental design, data collection, and analysis.
  • Self-E includes pre-configured experiments for users to choose from, while also supporting customizable experiments.
• Innovative Features:
  • Integration of traditional self-experimentation steps (e.g., intervention conditions, data input, analysis) to enable ordinary users to seamlessly execute experiments in daily environments.
  • Utilization of Bayesian techniques (Thompson Sampling) to intuitively analyze and present the likelihood and magnitude of causal relationships.
  • A user-friendly interface and framework designed to balance guidance for beginners and customization for advanced users.
• Implementation Steps and Key Technologies:
  1. Experiment Selection and Setup: Users first select or design experiments (variables for intervention and outcomes), including schedules, recording frequency, and quantifiable metrics.
  2. Data Collection: Experiments collect daily compliance or rating data from users.
  3. Real-Time Analysis: Self-E generates probability analyses based on recorded data (e.g., an intervention has a 74% positive probability of improving user happiness).
  4. Feedback and Modification: Users are guided to redesign experiments or accept existing results and take follow-up actions.

Research Outcomes

• Specific Outcomes:
  1. User Experience: Feedback from two study phases (including 16 community users and advanced experimental design users) was collected regarding Self-E.
  2. Scientific/Behavioral Pattern Discovery:
     • Ordinary users are often influenced by existing health behavior assumptions, which may actively or passively alter experimental settings.
     • Customizable experiment features are favored by advanced users, but their needs vary widely.
• Advantages Compared to Existing Solutions:
  • Compared to domain-specific programs like TummyTrials, Self-E offers a more universal cross-domain option.
  • Focus on balancing guidance and operability makes it suitable for ordinary users with no experimental design experience.
• Experimental or Evaluation Results:
  • An average compliance rate of 73% demonstrates the feasibility of short-term experimental strategies.
  • User feedback highlights the importance of promoting randomization and the potential value of data modification features.
• Limitations and Future Directions:
  • Limitations:
    • Small sample size: The recruited group primarily consisted of healthy university students, which may limit the generalizability of the results.
    • Time constraints: The two-week experiment duration is relatively short, making it difficult to capture long-term behavioral impacts.
  • Future Directions:
    • Enhance user education features: Provide more comprehensive explanations of the theoretical basis and methods of self-experimentation.
    • Socialized research: Explore the potential benefits of users sharing results or designing experiments collaboratively.
    • Improve personalization and controllability: For example, distinguishing "extreme data points" and enabling more complex experimental designs (multi-variable and multi-condition).

The authors provide practical insights for user-centered personal health technology development, ranging from studying how users utilize mobile applications for self-experimentation to discovering causal relationships in health behaviors and offering actionable improvement directions.