Gender and Sexuality Bias in Large Language Models: A Longstanding Issue at an Expanded Scale

1

Gender bias in large language models (LLMs) is the most consistently documented form of demographic bias affecting clinical documentation.
2

Approximately one-third of adults seek health information from AI chatbots, often due to financial obstacles and access barriers.
3

LLMs exhibit various biases, including gender, racial, and socioeconomic biases, which can exacerbate existing health inequities.
4

Clinical evaluations show that gender bias in LLMs can lead to divergent management recommendations based on sociodemographic descriptors.
5

Current AI evaluation frameworks often conflate sex and gender, complicating the assessment of clinically relevant differences.

1