미국의 의료 분야 차별금지법: 인공지능의 편향에 대한 대응과 시사점

Antidiscrimination law in U.S. healthcare began with the Civil Rights Act of 1964 and has since expanded under Section 1557 of the Patient Protection and Affordable Care Act. The protected characteristics under Section 1557 include race, color, national origin, sex, age, and disability. As in other sectors, machine learning-based algorithms present a challenge in applying antidiscrimination law due to their opaque, black-box nature. At the practical level, the tendency of many courts to deny private right of action in a disparate impact claim under Section 1557 acts as a disincentive for people in a protected class to seek legal action. In Korea, the challenges in preventing or mitigating algorithmic bias in healthcare are compounded. First and foremost, no antidiscrimination law exists in the healthcare sector yet. Notwithstanding, the artificial intelligence bills proposed to the National Assembly and the ethical guidelines published by the medical community stipulate that preventing discrimination or bias is essential in developing artificial intelligence, leaving in the vacuum which characteristics should be protected. In addition, health disparity has not been a significant issue in Korean healthcare in general. For Korea, the initial efforts should lay down the necessary first step by defining the protected characteristics that reflect public health research while being compatible with other sectoral antidiscrimination laws.