LLMs can bridge the gap between natural-language questions and rigorous causal analysis by mapping user queries to structured causal models, making complex statistical reasoning accessible to domain experts without deep technical expertise.
teLLMe is a system that helps traffic researchers answer causal questions about urban driving using dashcam video data. It combines causal structure learning with large language models to translate natural-language questions into structured causal analyses, returning explainable results that show estimated effects, assumptions, and uncertainty.