Ensuring Equity in AI: Addressing Data-Driven Biases in Health-specific Large Language Models for Global Use

Khalid S., Hasheminasab A., Eade J.

This blog post explores how biases in health-specific AI models trained on Global North data limit their effectiveness in South Asia. We found that fine-tuning on local patient records improves performance, highlighting the need for representative datasets and regional collaboration to build equitable, accurate AI tools for global healthcare.

DOI

10.21955/gatesopenres.1117193.1

Type

Other

Publication Date

2025-05-08T00:00:00+00:00

Permalink More information Close