Investigating Gender Stereotypes in Large Language Models via Social Determinants of Health

2026-03-10

AI Summary

This study addresses the risk of gender bias propagation in large language models (LLMs) within sensitive domains such as healthcare, where existing bias evaluation frameworks often overlook the interactions among social determinants of health (SDoH) and their contextual dependencies. For the first time, this work employs multidimensional SDoH interactions as probing constructs, integrating prompt engineering and controlled probing experiments to systematically analyze stereotypical behaviors in prominent LLMs when gender co-occurs with other SDoH factors in French clinical texts. The findings demonstrate that LLMs activate gendered stereotypes in response to SDoH-related inputs, thereby confirming that incorporating SDoH interactions significantly enhances and refines current bias assessment methodologies.

Technology Category

• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy
• Machine Learning: Ethics, Bias, and Fairness
• Cognitive Modeling & Cognitive Systems: Social Cognition And Interaction

Application Category

• Social Networks and Social Media: Fairness and bias in social network and social media analysis
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

Large Language Models (LLMs) excel in Natural Language Processing (NLP) tasks, but they often propagate biases embedded in their training data, which is potentially impactful in sensitive domains like healthcare. While existing benchmarks evaluate biases related to individual social determinants of health (SDoH) such as gender or ethnicity, they often overlook interactions between these factors and lack context-specific assessments. This study investigates bias in LLMs by probing the relationships between gender and other SDoH in French patient records. Through a series of experiments, we found that embedded stereotypes can be probed using SDoH input and that LLMs rely on embedded stereotypes to make gendered decisions, suggesting that evaluating interactions among SDoH factors could usefully complement existing approaches to assessing LLM performance and bias.

Problem

• gender stereotypes
• large language models
• social determinants of health
• bias propagation
• healthcare NLP

Innovation

• social determinants of health
• gender bias
• large language models
• bias probing
• healthcare NLP