Personalised Recommendations in Mental Health Apps: The Impact of Autonomy and Data Sharing

Title of the Paper

Research on Personalized Recommendations in Mental Health Applications: The Impact of Autonomy and Data Sharing

Bibliographic Information

• Subject Area: Digital Mental Health Applications, User Experience, Personalized Recommendations
• Keywords: User Perception, Personalization, Recommendation Systems, Personality Traits, Data Privacy, Autonomy, App Engagement, Data Sharing, Mental Health Apps, User Experience

Research Background and Problems

• Identified Issues or Challenges:

  1. The rapid growth of digital mental health applications has prompted researchers to explore how to effectively deliver personalized recommendations from the user's perspective, reducing content selection overload.
  2. Balancing personalized guidance with maintaining user autonomy in the design of personalized technologies.
  3. Data sharing is fundamental to achieving personalized recommendations, but privacy concerns pose challenges to user acceptance of data sharing.

• Importance:

  • Research shows that personalized health interventions can improve engagement and ultimately therapeutic outcomes.
  • User behavior and reliance on tools are critical factors for the success of digital health interventions. Understanding user preferences can better inform the design of experiences that encourage long-term engagement.

• Research Motivation and Related Work:

  • Previous studies focused on providing choices in health applications, but this research emphasizes users' perceptions of recommendation methods (autonomous choice vs. guided choice) and data sharing approaches (self-reported vs. smartphone data), as well as their impact on actual usage behavior.
  • A deeper exploration of the ethical tension between privacy and personalization.

Solution

• Methods and Solutions:

  • A placebo randomization-controlled trial was proposed, designing a two-factor experiment (autonomy and data sharing).
  • Tested two user experience models (autonomous choice vs. guided choice) and two data collection methods (completing a personality questionnaire vs. allowing smartphone data collection).

• Innovations:

  • Investigated the relationship between user preferences and actual usage behavior in mental health applications.
  • Systematically compared the impact of different design variables on user engagement and preferences.

• Implementation Steps and Key Techniques:

  1. Utilized a commercial mental health application with multiple activity modules.
  2. Designed a user experiment framework consisting of a pre-survey, a seven-day app usage period, and an exit survey.
  3. Conducted statistical analysis of experimental data to compare the effects of "autonomy" and "data sharing methods" on user preferences and behavior.
  4. Used randomized recommendation content (non-genuine personalization) to mitigate the impact of inconsistent recommendation model accuracy.

Research Outcomes

• Specific Findings:

  1. Impact of Autonomy:
     • User behavior: Applications offering primarily autonomous choice experiences were used three times more than those with guided choice.
     • User preferences: Although user behavior favored autonomy, surveys indicated that most users preferred applications to provide activity recommendations.
  2. Impact of Data Sharing Methods:
     • User sharing behavior: Users were more willing to report their personality through questionnaires rather than allowing smartphone data collection.
     • Privacy risk perception: Users generally perceived greater privacy risks in allowing mental health apps to access smartphone sensory data.
  3. Personalized Recommendations and App Engagement:
     • Behavioral data indicated no significant difference in app usage based on the source of recommendations (automated data inference vs. questionnaire).

• Advantages Compared to Existing Solutions:

  • Provided quantitative design recommendations for mental health applications by analyzing both user behavior and preferences.
  • Addressed the conflict and balance between data privacy and personalization in user experience design.

• Experimental or Evaluation Results:

  • The number of activities completed by the autonomy group was significantly higher than that of the guided group. Overall, users preferred autonomy-focused designs but still desired some recommendation elements.
  • The questionnaire-based method for inferring personalized user models was more trusted by users than automated smartphone data-based models.

• Limitations and Future Directions:

  • Limitations:
    1. The design of randomized recommendations limited the exploration of the relationship between accuracy and perceived personalization.
    2. The study was conducted only in Europe, potentially introducing cultural bias.
    3. The lack of actual smartphone data collection reduced opportunities for in-depth analysis of individual behavioral characteristics.
  • Future Directions:
    • Develop genuine personalized recommendations and conduct experiments combining user preferences and behaviors.
    • Explore the impact of personalized design on long-term app engagement and therapeutic outcomes.

Conclusion

This study provides practical guidance for designers of mental health applications. The findings suggest that delivering personalized recommendations should strike a balance between enhancing user experience, fostering a sense of autonomy, and protecting privacy.