Impracticality of banning collection of data on ethnicity and race in artificial intelligence-enabled health care in France.

Abstract

Artificial intelligence (AI) or machine learning (ML) are revolutionising health care, enhancing diagnostics and treatment through AI-enabled medical devices. However, the effectiveness of AI or ML models is hindered by substantial biases, particularly against minority populations such as African and Afro-descendant people. In this Viewpoint, we discuss the colour-blind policy of France, which prohibits the collection of data on race and ethnicity, and so ignores racial distinctions and consequently does not address such biases inherent to AI models. France's historical stance on non-racial distinction, aimed at promoting equality, paradoxically conceals the racial disparities embedded in AI technologies. This paradox is particularly evident in health care as AI's capability to discern race from health data challenges the effectiveness of colour-blind policies. By examining the implications of France's colour-blind policy and the necessity of race-conscious strategies, we advocate for a fundamental shift towards incorporating ethnic and racial data in AI development. This shift will not only enhance the accuracy and fairness of AI applications in health care but also ensure that France can compete internationally in the AI-enabled medical device market while upholding ethical AI practices that prioritise transparency and equity. Urgent policy revisions are required to integrate race and ethnicity considerations across the entire lifecycle of AI, including training data used for AI models in health care, thereby improving health outcomes for all population groups, including marginalised communities.