Negative perceptions of mental illness and other health issues subtly shape outputs from large language models

Read original article