AI, Health Equity and Ethical Implementation: Webinar Takeaways

The second webinar in a three‑part series examined how artificial intelligence (AI) intersects with health equity and ethical practice. Dr. Yao Haye and Jen Antilla led the session, building on last week’s AI primer and previewing next week’s focus on AI‑driven lab management systems. The Digital Initiatives group at ITC, which runs health information exchange and workforce development projects, hosted the discussion.

AI and Health Equity

Integrating AI into health care confronts several practical hurdles: upgrading infrastructure, closing knowledge gaps among professionals, overcoming resistance to change, and keeping pace with rapid AI evolution. Within this context, health equity is defined as the fair and just distribution of AI‑powered health technologies and their benefits, ensuring high‑quality care regardless of socioeconomic status, race, sex, gender, ethnicity, disability, or location.

AI bias is described as an AI system designed in a way that makes its decisions or use unfair. Five bias types were highlighted:

• Experience and expertise bias – skew from varying expertise among developers and users.
• Exclusion bias – systematic under‑representation of certain groups in training data.
• Environment bias – omission of social determinants of health or contextual factors in data collection.
• Empathy bias – difficulty embedding human experiences and subjective elements.
• Evidence bias – distortion in generating, disseminating, or translating scientific evidence.

Addressing these biases calls for diverse expert teams, continuous training, inclusive data collection, equity audits, incorporation of social determinants of health and qualitative data, and diversified funding streams.

Ethical and Responsible AI Implementation

AI can amplify existing societal biases, so its deployment should aim to reduce, not widen, health disparities. Reported harms include large language models suggesting dangerous actions, false diagnoses eroding trust, profiling patients as likely to default on treatment, cyber‑attacks leading to data breaches, misleading health information from chatbots, and misdiagnosis or delayed treatment such as missed cancers.

The World Health Organization (WHO) proposes six core principles for ethical AI:

1. Protect autonomy – ensure human oversight, informed consent, and privacy.
2. Promote human well‑being and safety – minimize harm, maximize benefits, and monitor AI performance.
3. Ensure transparency and explainability – document design assumptions, data sources, and processes; conduct audits.
4. Foster responsibility and accountability – clarify roles, define responsibilities, and establish redress mechanisms.
5. Ensure inclusiveness and equity – mitigate bias, involve diverse teams and data, guarantee equitable access.
6. Promote responsiveness and sustainability – continuously evaluate and update systems while minimizing environmental and workforce impacts.

Ethical frameworks integrate these principles from project inception, emphasizing bias mitigation, ethical data use, interdisciplinary collaboration, equity audits, community engagement, and public education. Embedding equity into machine‑learning models requires attention to equity and ethics throughout design, data collection, training, and validation.

Discussion and Challenges

Participants identified pressing challenges: limited understanding of AI capabilities, lagging governance and policy, faster AI advancement than equity discussions, difficulty selecting appropriate use cases, patient consent for data use, unclear accountability for AI outcomes, scarce data‑sharing agreements, minimal AI use in clinical decision‑making, under‑representation of health professionals in AI institutes, practical implementation of human‑in‑the‑loop, and the need to educate primary‑care workers on AI systems.

Stakeholder roles were outlined: healthcare experts and data scientists on vetting boards; multi‑sector policymakers for a coordinated approach; developers who grasp data representativeness and ethical issues; organizational leaders driving change management; clinicians and public‑health professionals defining useful tools and participating in design; policymakers enforcing AI governance; and providers and patients educating themselves on risks and ethical principles.

Conclusion and Next Steps

Balancing AI innovation with oversight, transparency, and accountability is essential. Guidelines should ensure AI augments, rather than replaces, human judgment. Involving diverse stakeholders—communities, patients, experts, and developers—throughout the AI lifecycle builds transparency, trust, and acceptance. Clarifying responsibilities and liabilities, establishing institutional policies, and providing performance oversight are critical next steps. The upcoming third webinar will demonstrate AI applications in lab management systems, and resources from earlier sessions will be shared.