When AIs outperform doctors: confronting the challenges of a tort-induced over-reliance on machine learning
A. Michael Froomkin, Ian Kerr, Joëlle Pineau
Abstract
Machine learning (ML) diagnostics will surpass human doctors in accuracy, a shift that poses significant implications for medical malpractice law, the provisioning of medical services, the demand for human doctors, and the quality of medical diagnostics. Once ML diagnostics are recognized as superior, they will become the standard of care, yet this transition could paradoxically undermine patient safety standards. Initially, combining human and machine diagnostics may offer superior outcomes by diversifying error types. However, as ML systems improve, there may be overwhelming legal and ethical pressures to fully delegate diagnostics to machines. This progression could lead to future editions of ML systems being trained on ML-generated data, making it hard to ensure predictive quality. Medical liability laws may need reform to avoid negative diagnostic quality trends. Malpractice rules, or other economic incentives, may need adjustments to keep human physicians in the loop, thus ensuring the maintenance of care quality.