New AI Models Still Reproduce Racial And Gender Stereotypes In Medicine

Researchers at Flinders University found that current AI models continue to reflect historical biases by linking specific diseases and health outcomes to certain racial and gender groups. These findings highlight that despite technological advancements, developers have yet to effectively filter out societal prejudices from training datasets. This persistence of bias poses a significant risk to patient equity, as clinical tools relying on these models may perpetuate health disparities by providing inaccurate or prejudiced diagnostic recommendations.