Physician Liability Impacts AI Algorithm Design and Use

The increasing use of AI in clinical decision-making has raised concerns about algorithmic disparities, where a single algorithm may perform unequally across different patient groups. In response, liability rules have emerged, holding healthcare providers responsible when their reliance on such disparate algorithms leads to erroneous decisions. This research explores the economic implications of these liability considerations on both AI firms and physicians. The study models an AI firm designing an algorithm for two patient groups, where improving accuracy for a disadvantaged group is more costly. Physicians, as accountable decision-makers, then weigh the benefits of AI consultation against potential liability. The findings suggest that liability rules can lead to unequal AI adoption, with physicians potentially reducing overall AI use or using it less for disadvantaged patients. Paradoxically, mandating equal algorithmic accuracy might harm both groups by distorting the firm's investment incentives and the physician's AI-use decisions.