First, Do No Harm (With LLMs): Mitigating Racial Bias via Agentic Workflows

TL;DR

This study evaluates racial bias in five medical LLMs across two clinical tasks, finding that agentic workflows can reduce explicit bias and emphasizing the importance of multi-metric bias assessment in healthcare AI.

Contribution

It introduces a structured evaluation framework for racial bias in medical LLMs and demonstrates that agentic workflows can mitigate some biases in diagnostic tasks.

Findings

GPT-4.1 showed the smallest deviation from racial distributions in synthetic tasks.

DeepSeek V3 achieved the best overall results in differential diagnosis.

Agentic workflows improved bias metrics, though not uniformly across all metrics.

Abstract

Large language models (LLMs) are increasingly used in clinical settings, raising concerns about racial bias in both generated medical text and clinical reasoning. Existing studies have identified bias in medical LLMs, but many focus on single models and give less attention to mitigation. This study uses the EU AI Act as a governance lens to evaluate five widely used LLMs across two tasks, namely synthetic patient-case generation and differential diagnosis ranking. Using race-stratified epidemiological distributions in the United States and expert differential diagnosis lists as benchmarks, we apply structured prompt templates and a two-part evaluation design to examine implicit and explicit racial bias. All models deviated from observed racial distributions in the synthetic case generation task, with GPT-4.1 showing the smallest overall deviation. In the differential diagnosis task, DeepSeek V3 produced the strongest overall results across the reported metrics. When embedded in an agentic workflow, DeepSeek V3 showed an improvement of 0.0348 in mean p-value, 0.1166 in median p-value, and 0.0949 in mean difference relative to the standalone model, although improvement was not uniform across every metric. These findings support multi-metric bias evaluation for AI systems used in medical settings and suggest that retrieval-based agentic workflows may reduce some forms of explicit bias in benchmarked diagnostic tasks. Detailed prompt templates, experimental datasets, and code pipelines are available on our GitHub.