Assessing Changes in Healthcare Outcomes After Artificial Intelligence Implementation

Abstract

Generative techniques in artificial intelligence (AI)-and large language models in particular-have proliferated in recent months and years. Automation of traditionally human tasks introduces efficiencies, sometimes radically reducing the time required for certain jobs or eliminating them altogether. But the consequences of automation are uneven. For AI to be a net positive for society, it must enhance social productivity and push the frontier of production and services further outward. In these discussions, hyperbole frequently overshadows calm reflection about AI’s broader impact, whether for good or ill. Some parties have warned that this new technology is not only fundamentally altering how humans create, work, and interact but, unless restricted, poses a risk not just to the future of work but also to human civilization itself. Others view it as an opportunity to reach previously unattainable levels of productivity and social and human progress. A balanced perspective demands asking how AI is affecting the delivery of health care - a domain in which many processes can theoretically be automated, providing opportunities for exceeding previous levels of productivity and human experience. Until recently, the solution to such questions lay mostly in speculation. Indeed, conversations about AI and health care almost always revolved around technology and machine learning development. Yet clinicians, care teams, and hospitals worldwide are now using AI tools and, in some cases, responding to them as virtual team members. As this use of AI tools in health care spreads, it becomes possible to compare pre-AI and post-AI situations. In principle, such comparisons should focus on the predefined substantive impact framework of health care delivery rather than on speculating about AI’s potential influence on future delivery.