Promise or Peril? Exploring Black Adults' Perspectives on the Use of Artificial Intelligence in Health Contexts

Paper Title

Promise or Peril? Exploring Black Adults' Perspectives on the Use of Artificial Intelligence in Health Contexts

Publication Info

• Topic area: Black adults' perspectives on AI in healthcare and its implications for health equity.
• Keywords: Artificial intelligence, health equity, racial bias, Black communities, participatory design, healthcare innovation, health disparities, mental health, structural barriers, qualitative research.

Background and Problem

• Problem / challenge: Limited research exists on Black communities' perspectives regarding AI in healthcare, despite their experiences with significant health inequities and the potential for AI to perpetuate or mitigate these disparities.
• Significance: Understanding these perspectives is crucial for designing equitable AI systems that address structural barriers and health inequities affecting Black populations.
• Motivation and related work: Previous studies have documented racial bias in healthcare and AI systems, as well as the importance of participatory methods in technology design. However, there is a lack of comprehensive research on Black adults' attitudes toward health AI, particularly its role in addressing health equity.

Solution

• Proposed approach: A mixed-methods study involving surveys and qualitative workshops with 18 Black adults to explore their perspectives on health AI and its potential to support health equity.
• Novelty:
  1. Characterization of nuanced attitudes among Black adults toward health AI, ranging from optimism to skepticism.
  2. Identification of structural barriers to health equity and how participants perceive AI's role in addressing or exacerbating these barriers.
  3. Introduction of the concept of health AI as "armor" to protect against systemic healthcare biases.
  4. Framework for addressing intended, unintended, and inevitable harms in health AI design.
• Procedure and key techniques:
  • Conducted pre-workshop surveys to assess attitudes toward AI and health equity.
  • Held three in-person workshops with participants to discuss health AI scenarios, survey results, and structural barriers to health equity.
  • Used thematic analysis to identify key themes from 16.5 hours of workshop discussions.

Results

• Concrete findings:
  • 61% of participants believed AI would improve health outcomes, higher than the 38% reported in a national survey.
  • 83% viewed racial bias in healthcare as a major problem, compared to 35% nationally.
  • Participants expressed mixed feelings about AI's role in pain management, mental health chatbots, and skin cancer screening.
  • AI was seen as a potential tool to counter healthcare biases and improve access but also as a source of potential harm due to biases in its design and implementation.
• Advantage over baselines:
  • Participants were more optimistic about AI's potential to address health inequities compared to national survey respondents.
  • Greater emphasis on the structural and community-oriented implications of health AI.
• Experiments / evaluation:
  • Workshops explored participants' attitudes toward AI in specific health scenarios (e.g., pain management, mental health, skin cancer screening).
  • Survey results were compared to national data to highlight differences in perspectives.
• Limitations and future work:
  • The study focused on a single metropolitan area in the Southeastern U.S. with a small sample size (n=18), limiting generalizability.
  • Future research should explore the intersection of race, sexual orientation, and other identities in shaping health AI attitudes.
  • Partnerships between community organizations, healthcare providers, and technology companies are needed to translate findings into practice.

Summary

This study explores Black adults' perspectives on the use of AI in healthcare, highlighting both optimism for its potential to address health inequities and concerns about its risks. Participants viewed AI as a tool that could counter healthcare biases and improve access but also expressed skepticism about its maturity and potential to perpetuate systemic inequities. The findings emphasize the importance of engaging marginalized communities in health AI design and propose a framework for addressing intended, unintended, and inevitable harms. Future work should focus on broader, intersectional research and fostering collaborations to ensure equitable health AI innovation.