Have we been measuring AI political bias wrong? A better approach is possible

This blog post by David Rozado explores the differences between ideological preferences and epistemic failures in large language models (LLMs) and argues for a more nuanced approach to measuring political bias in AI systems.