Towards a Multi-Stakeholder process for developing responsible AI governance in consumer health

Area of Science:

• Health Informatics
• Artificial Intelligence in Healthcare
• AI Governance

Background:

• Rapid advancement of AI in healthcare presents both opportunities and risks.
• Current AI governance lacks domain-specific recommendations for consumer health.
• Generic guidelines may lead to unintended consequences in diverse healthcare contexts.

Purpose of the Study:

• To explore pragmatic, multi-stakeholder governance approaches for consumer-facing health AI.
• To establish a tailored governance framework for consumer health AI.
• To identify key constraints and desirable characteristics for health AI governance models.

Main Methods:

• A 4-month multidisciplinary expert consensus process involving nearly 200 participants.
• Synthesis of insights to guide the development of proposed governance models.
• Focus on collaborative deliberation to inform practical solutions.

Main Results:

• A consensus emphasizes limiting scope and incorporating multi-stakeholder, patient-centered perspectives.
• Desirable governance characteristics include adaptability, patient empowerment, and transparency.
• A recommended pathway involves forming a Health AI Consumer Consortium (HAIC2) to align incentives.

Conclusions:

• Core themes of consumer voice, transparency, and balanced oversight are globally relevant.
• Proposed principles for multi-stakeholder alignment and patient empowerment can ensure AI is safe, effective, equitable, and trustworthy (SEET).
• This framework offers a productive approach for governing AI in consumer health worldwide.