As artificial intelligence (AI) tools increasingly inform medical decisions, experts caution that these systems cannot fully account for individual patient values, priorities, and circumstances—factors integral to many healthcare choices. Physicians and researchers emphasize that while AI can process vast amounts of clinical data rapidly, it lacks the ability to understand the personal context that shapes treatment decisions.
According to James N. Weinstein, a surgeon and health technology executive, and Ogan Gurel, a physician and AI researcher, many common medical conditions do not have a single clearly “right” answer. Instead, successful outcomes often depend on how well a chosen treatment aligns with the patient’s goals, tolerance for risk, and lifestyle. For example, individuals with back pain may face similar clinical findings on diagnostic scans but differ in their preferences—some may prioritize a quick return to physically demanding work even if it requires surgery, while others may prefer to avoid operations despite longer recovery times.
The limitations of AI become especially clear in situations where medical options have comparable risks and benefits, such as decisions about treatment for low-risk prostate cancer, management of atrial fibrillation, or choices between surgery and physical therapy for chronic pain. AI systems are adept at calculating probabilities and suggesting the most common outcomes based on demographic and medical data, but they cannot grasp what a given patient values most, nor do they understand the nuances of physician-patient dialogue where facts, concerns, and trust converge.
In one case cited by Weinstein and Gurel, an AI-driven symptom checker flagged a heart rhythm abnormality in a retired teacher and recommended an invasive procedure. However, after discussing the matter, physicians determined that the patient’s priority was to avoid prolonged recovery in order to maintain travel plans and family visits. Medication and monitoring, less dramatic but supported by evidence, better suited these goals. This example highlights the discrepancy between AI-generated recommendations and personalized care.
The authors note that patient preferences significantly impact approximately 25 percent of healthcare spending in the United States. Ignoring these preferences—whether by clinicians or algorithms—can lead to misaligned care, unnecessary procedures, higher costs, poor treatment adherence, and patient regret.
They advise patients using AI-based health tools to ask critical questions: “Best for whom?” to clarify if a recommendation is general or tailored; “What does this system not know about me?” recognizing AI’s inability to perceive personal circumstances; and “What happens if I wait or choose differently?” reminding patients that many decisions are not urgent and merit reflection.
Ultimately, the experts assert, AI should serve as a support tool to inform human judgment rather than replace it. While AI can integrate extensive medical knowledge, it remains limited in appreciating the individual experiences and values that shape the best healthcare decisions. The most important unknown variable in medicine, they argue, is the patient themselves.