Participatory AI Considerations for Advancing Racial Health Equity

Research Background and Issues

• Identified Problems or Challenges:
  The authors observed that health-related artificial intelligence systems (Health AI) are rapidly advancing, yet the design process of these systems often lacks the participation of racially marginalized groups. This poses a significant issue for racial groups that have long suffered from health inequities due to social and structural factors, as poorly designed systems could further exacerbate their health conditions.

• Significance:
  The issue of racial health inequities holds profound societal importance, as it pertains to life safety and overall health and well-being. If these issues are not adequately addressed in Health AI, these technologies may simultaneously improve health outcomes while intensifying existing inequities.

• Research Motivation and Related Work:
  The authors reviewed the current racial health inequity issues in Health AI design, highlighting the limited involvement of racially marginalized groups in the design process and the absence of a systematic practice framework to guide fair and effective participation. Additionally, they reflected on existing participatory AI design research and identified several limitations, such as the lack of comprehensive discussion on health inequities.

Proposed Solutions

• Proposed Strategies:
  The authors proposed eight strategies to address the limitations of existing research, including:

  1. Community-centered design approaches.
  2. Supporting collaborative learning to counter symbolic representation.
  3. Comprehensive community participation from high-level to low-level design.
  4. Encouraging design thinking about power dynamics in Health AI.
  5. Promoting goals beyond survival to foster prosperity among racial groups.
  6. Exploring the role of emotions in Health AI design.
  7. Embracing collectivist perspectives through relational approaches.
  8. Supporting long-term community-based design participation.

• Innovative Contributions:
  The authors expanded existing participatory AI design frameworks, emphasizing that "participation" should not be limited to data collection and model design but should encompass holistic planning, policy-making, and sustained community involvement. Furthermore, they introduced the Health Equity Framework to provide theoretical guidance for AI designs that support racial health equity.

• Implementation Steps and Key Techniques:

  1. Employ purposeful sampling and community-centered methods to ensure diverse perspectives.
  2. Bridge participants' knowledge gaps through collaborative learning.
  3. Support multi-level design participation, including high-level design (macro societal issues), mid-level design (AI system goals and functionalities), and low-level design (datasets and models).
  4. Integrate discussions on power, emotions, and relationships into design thinking to address complex societal challenges.

Research Outcomes

• Specific Outcomes:
  The authors identified eight major research gaps related to racial health equity in participatory Health AI design and proposed corresponding strategies for addressing these gaps. Additionally, they highlighted critical themes—such as emotions, power, prosperity, and relationships—as "axis factors" for promoting racial health equity.

• Advantages Over Existing Solutions:
  Compared to existing participatory AI frameworks, this study places greater emphasis on the unique needs of racial health equity and proposes design concepts that integrate community contexts and complex social structures. Moreover, the recommendations focus on fundamentally addressing health inequities rather than merely making superficial adjustments to symptoms.

• Experimental or Evaluation Results:
  This paper primarily focuses on theoretical analysis and does not involve experiments or practical evaluations. Conclusions were drawn through literature analysis, and future directions were proposed.

• Limitations and Future Directions:
  The study's limitations include the absence of specific case studies or empirical research to test the practical effectiveness of the strategies. Additionally, the racial health equity framework is primarily tailored to the U.S. context, necessitating further testing and validation in diverse cultural environments. Future research should also develop tools for effectively evaluating the impact of these strategies, such as metrics for community knowledge and participation satisfaction, as well as standards for assessing the health equity of designs.

Conclusion

This paper provides an extended theoretical framework for participatory Health AI, focusing on racial health equity issues, with significant academic and practical implications. The authors emphasize that designing Health AI must transcend existing isolated approaches, foster inter-community collaboration, respect diversity, and advance lasting change through collaborative learning. This study offers critical insights for future research and represents an important direction in pursuing social justice and technological fairness.