Unmasking and Quantifying Racial Bias of Large Language Models in Medical Report Generation

Large language models like GPT-3.5-turbo and GPT-4 hold promise forhealthcare professionals, but they may inadvertently inherit biases duringtheir training, potentially affecting their utility in medical applications.Despite few attempts in the past, the precise impact and extent of these biasesremain uncertain. Through both qualitative and quantitative analyses, we findthat these models tend to project higher costs and longer hospitalizations forWhite populations and exhibit optimistic views in challenging medical scenarioswith much higher survival rates. These biases, which mirror real-worldhealthcare disparities, are evident in the generation of patient backgrounds,the association of specific diseases with certain races, and disparities intreatment recommendations, etc. Our findings underscore the critical need forfuture research to address and mitigate biases in language models, especiallyin critical healthcare applications, to ensure fair and accurate outcomes forall patients.

Further reading#

• Access Paper in arXiv.org