Digital Phenotyping as Felt Informatics: Designing AI-Based Mental Health Diagnostic Tools Through Aesthetics

Research Background and Issues

• Issues and Challenges: Mental health diagnostic tools lag behind those in other medical fields, relying primarily on subjective reports and clinical interviews, which are neither objective nor precise. Digital Phenotyping (DP) has been proposed as a method to collect data using smart devices and wearables, aiming to address the shortcomings of current diagnostic models through objective data. However, existing DP designs often overlook patient needs, reducing the complex mental health experiences of users to quantifiable data models. This simplification raises ethical concerns, usage dilemmas, and marginalizes patient experiences.
• Significance: Accurate and objective diagnostic tools can bridge the gap between mental health patients and physicians, providing support for monitoring and treatment planning. However, such innovations may also exacerbate the neglect of subjective experiences while reinforcing stigmatization and societal labeling of mental health issues.
• Research Motivation and Related Work: The authors examine the hypothetical limitations of DP being positioned as a scientific and objective tool, exploring how current designs compress users' actual psychological and physical experiences and lack sensitivity to users' emotional and bodily experiences in daily life. Related work focuses on the ethical, sociological, and technological potential of DP, while this study aims to offer a new perspective centered on aesthetics and "Felt Informatics" to address these gaps.

Solution

• Proposed Approach: The authors reframe DP as a "Felt Informatics" issue, emphasizing the aesthetics of user "feeling" and AI "sensing." The goal is not merely to create AI that "understands" users but to design systems that allow users to "feel understood."
• Innovative Contributions:
  • Introduces the concept of "Felt Informatics," reexamining how DP captures, represents, and influences users' mental health experiences through an aesthetic lens, emphasizing the artistic interaction between data design and users' bodily perception and subjective experiences.
  • Advocates for transforming AI from a mere data analyst into an "Aesthetic Agent" capable of sensitively identifying underlying issues and collaborating with users to perceive and understand their mental states.
• Implementation Steps:
  1. Perception: Redefine the essence of perceptual data, focusing on DP's "hyperaesthetisation" and its impact on the selection and influence of user behavioral patterns.
  2. Representation: Improve data visualization design; avoid trends of labeling and pathologizing, and empower users with greater interactivity and reflective capabilities regarding their data.
  3. Experience: Explore the everyday manifestations of users' mental health, designing techniques and tools to guide individuals in reflecting on their state rather than directly deriving conclusions or diagnostic labels.
  4. Relationality: Promote the integration of DP systems into broader ecosystems of care, ensuring appropriate roles for families, communities, and physicians in the care relationship, rather than limiting interactions to just the user and technology.

Research Outcomes

• Specific Results:
  • Proposes a "Felt Informatics" design framework to guide the design of digital mental health applications, better aligning with users' psychological, sensory, and cultural contexts.
  • Provides four key design dimensions (Perception, Representation, Experience, Relationality) and several practical design questions to help designers create more sensitive and humanized digital medical tools.
• Comparison with Existing Solutions:
  • Current DP tools focus heavily on data collection and analysis while neglecting users' feelings and understanding of the data. The "Felt Informatics" framework elevates user experience and interaction to a central position.
  • Emphasizes the emotional effects and reflective space generated by data, rather than merely representing scientific data or diagnostic results.
• Experimental and Evaluation Results: No large-scale empirical data or user feedback is currently available; the study primarily focuses on theoretical exploration and design strategies within the framework.
• Limitations and Future Directions:
  • Limitations: The practical value of the aesthetic perspective for long-term health improvement and its applicability in complex medical scenarios require further empirical research. The specific methods for integrating with existing medical systems and standards remain unclear.
  • Future Directions:
    • Develop specific case studies or applications to validate the effectiveness of the "Felt Informatics" framework in practice.
    • Explore how "Felt Informatics" can be integrated into mental health ecosystems across different cultural contexts.
    • Enhance collaboration between AI and users to optimize perceptual synergy, improving model transparency and user trust.