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.