8
CONCLUSION

In the past several years, along with the rapid advances in artificial intelligence (AI) methods and technologies, there has been a concomitant spread of novel tools to support health, health care, and biomedical science. These AI tools offer the potential for transformational change in the prevention, detection, diagnosis, and treatment of disease. However, the fundamental differences in these technologies, relative to traditional health information technology, present new risks that require new governance frameworks and new approaches to learning. To that end, the National Academy of Medicine (NAM), under the direction of the AI Code of Conduct (AICC) steering committee, sought to align the health field and to catalyze collective action to ensure that the transformative potential of health AI is realized. The AICC’s Code Principles are mapped to the NAM’s Learning Health System (LHS) Shared Commitments and a set of simple rules, the Code Commitments, have been developed to aid in the employment of the principles in daily practice. With these aligned vision and goals, as powerful AI technologies and tools advance, the realization of the long-envisioned LHS is more imminently in reach.

However, AI can be likened to another powerful tool, a hammer, which embodies the duality of utility and intent. The hammer can be used to construct a school, a hospital, a cathedral—each the fulfillment of the collective aspirations of a community. Yet, the same implement can be employed to force incongruity, as in driving a square peg into a round hole, or to engage in a nefarious act, as in breaking a window to commit larceny. So it is with AI, that the value of the technologies and tools are defined by the purpose for which they are directed and governed. Here, the AI Code Principles and Code Commitments offer guideposts.

While health AI is on the cusp of supporting major breakthroughs in biomedical research, medical diagnosis and treatment, maintenance of health, and addressing health workforce challenges, much work remains to ensure that it meets its

potential. To meet the moment, all key stakeholders must play a role in ensuring that health AI advances humanity and avoids the risks associated with incongruent or malicious use of the tools and technologies. Ongoing collaboration across sectors is required to articulate the collective aspiration of the community and to advance AI governance. Components of that shared responsibility include:

• Promoting ongoing local, state, national, and international alignment of health AI governance frameworks, including the AICC framework;
• Applying and adapting the AICC Code Principles and Code Commitments to the local context;
• Translating the AICC framework into sector-specific implementation guides;
• Identifying and testing key use cases for application of the AICC framework;
• Establishing local, state, and national policies to support identified priorities; and
• Preparing the workforce to succeed in an AI-enabled environment.

While many technological advances have had potential to improve human health and well-being, they have failed to fully deliver on their promise. Likewise, the realization of a national LHS at scale has remained elusive, in part because contemporary U.S. health care is marked by misaligned incentives, fragmentation, discontinuity, and inundation. Intelligent applications of AI and collaborative governance can serve to bridge many of these gaps and bolster the sociotechnical infrastructure, moving away from fragmented and inchoate data to a system of cohesive and actionable knowledge. The AI Code Principles and Code Commitments can serve as guideposts for decision making about the application of AI tools, and more broadly, prioritizing these fundamental commitments can serve as a framework for increasing trustworthiness, enhancing organizational culture and synergy, accelerating discovery, and promoting effective care. Taken together, advances in health AI delivered in the context of shared principles and commitments can both fulfill the needs expressed in the earliest articulations of the LHS—continuous learning through a dynamic approach to evidence generation and improvement—and, most importantly, realize the shared goal of the health system writ large, to reduce suffering and create health for individuals, communities, and the population as a whole.