Tag
This study demonstrates that large language models inherit and amplify biases from stigmatizing language in clinical notes, leading to less aggressive patient management, and that current mitigation strategies are insufficient.