• AI Agents, Automaticity, and Value Alignment in Health Care

    R. Andrew Taylor, M.D., M.H.S. 

    Abstract

    As complex, largely autonomous tools capable of once inconceivable feats, such as building websites from simple directives, and guided by high-level, goal-oriented instructions, AI agents most likely represent the next frontier in the rapidly evolving landscape of health care AI. However, this significant shift from task-specific, narrowly focused traditional AI tools introduces substantial ethical and operational challenges. The autonomy and complexity of these agents demand an intentional shift of focus toward explicit value alignment that balances the diverse priorities of patients, clinicians, and society. To navigate these challenges, we advocate for an ethical system rooted in inclusivity, transparency, and adaptive governance, underpinned by robust stakeholder engagement, value elicitation, ethics-by-design integration, dynamic recalibration, and safeguards against misalignment. Embedding these principles into the design and deployment of AI agents, while maintaining accountability through transparent governance, is essential to ensuring that these technologies advance equity, enhance quality, and build trust in care.