Causal Bias Detection in Generative Artificial Intelligence

In three linesarXiv paper proposing a theoretical framework for detecting causal bias in generative AI models. Authors formalize causal fairness specific to generative models (vs standard ML), derive causal decompositions to quantify bias impacts across different causal pathways, and demonstrate their methodology by analyzing race and gender bias in large language models.

Summary generated by Claude — human-verified