Medical AI's Cost-Benefit Dilemma
A huge question is who medical AI tools are working for.
TL;DR
- Clinical AI tools in healthcare present an ethical challenge regarding whose interests they serve (patient vs. cost).
- Decision support tools face the same complex medical choices as human providers, requiring a balance between individual care and system costs.
- AI algorithms could influence decisions on diagnostic tests and expensive treatments, concentrating power in the hands of designers.
- Existing health chatbots are designed for engagement, not necessarily patient well-being.
- Wachter proposes transparency, with AI offering multiple recommendations and outlining tradeoffs.
- Lawsuits against UnitedHealth and Humana allege AI tools were used improperly to determine the end of post-acute care or recovery times.
- Past research showed a hospital algorithm prioritizing spending overlooked high-risk Black patients, leading to adjustments for racial bias.
- Traditional legal frameworks like malpractice and product liability are ill-suited for AI failures due to variable performance and opacity.
- Hospitals may be tasked with vetting AI models but often lack the necessary information.
- The issue is expected to generate significant political debate and litigation.