Artificial intelligence (AI) holds the promise of helping physicians make better treatment plans for their patients’ unique needs. While doctors now have more access to AI than ever, with hundreds of AI-based medical devices approved by the FDA, the technology’s full potential has yet to be realized. Tinglong Dai, PhD, and Shubhranshu Singh, PhD, two professors from the Johns Hopkins Carey Business School and core faculty members in the Hopkins Business of Health Initiative, authored and posted a paper online examining how physicians could decide to use AI.
Dai and Singh developed a theoretical model demonstrating that physicians’ use of AI in their clinical practice could be influenced by their liabilities to their patients and revenue generated from its use. Physicians could have the greatest incentive to use AI when they are confident in their treatment decision and expecting low liability. Overutilizing AI in these scenarios could raise revenues, which are beneficial for the practice though not for keeping healthcare costs down.
In the paper, Dai and Singh offer an unexpected perspective that differs from policymakers and AI developers. Whereas the latter feel enthusiastic that as AI becomes more precise it will be more widely used and replace the standard of care, the researchers suppose that physicians could feel apprehensive of the pressure AI puts on their decision-making when they are at odds with one other. For example, there could be an uncertain situation where a physician’s expertise and experience deem it best to prescribe a higher chemotherapy dose than the routine regimen. If this plan deviates from the AI’s treatment prediction, the physician may feel the burden of higher liability if the outcome proves to be unfavorable. The potential for conflict with the AI could dissuade the physician from using it in such a case. As such, the researchers suggest that "policymakers should focus on incentivizing physicians to use AI appropriately but not necessarily more widely."