Building Fair and Trustworthy Biomedical AI: A Tool for Identifying Key Decision Points

Area of Science:

• Biomedical AI
• Health Informatics
• AI Ethics

Background:

• Artificial intelligence (AI) offers transformative potential in biomedicine, improving diagnostics, drug discovery, and patient care.
• However, AI implementation raises significant ethical concerns, including algorithmic bias, health inequities, and data privacy risks.
• Addressing these requires a focus on fairness, trust, and trustworthiness in AI development and deployment.

Purpose of the Study:

• To propose a framework for embedding ethical responsibility in biomedical AI.
• To introduce the Trustworthy AI Decision Map, an adapted tool to guide ethical decision-making across the AI lifecycle.
• To facilitate multi-stakeholder dialogue and institutional accountability in developing fair and trustworthy AI.

Main Methods:

• Adapted a decision-mapping framework from precision medicine research.
• Developed the Trustworthy AI Decision Map to identify key ethical decision points in AI development.
• Illustrated the tool's application using a case study of AI deployment in rural healthcare.

Main Results:

• The Trustworthy AI Decision Map visualizes critical decision points impacting AI fairness and trustworthiness.
• The map aids in anticipating downstream consequences and integrating diverse stakeholder perspectives.
• It supports the establishment of institutional accountability for ethical AI practices.

Conclusions:

• Ethical considerations must be integrated at institutional and individual levels in biomedical AI.
• Multi-stakeholder engagement, particularly with underrepresented groups, is crucial for equitable AI.
• Further empirical validation of the Trustworthy AI Decision Map is needed to refine its utility in promoting fair and trustworthy biomedical AI.