CASTCurate: An Agentic System to Accelerate the Collection and Annotation of Data-Driven Stories

Abstract

This study introduces an AI-powered data storytelling agent designed to support data science educators by automatically curating high-quality, real-world data stories. The system streamlines the discovery of relevant instructional examples for specific teaching activities, including assignments, quizzes, classroom discussions, and case studies, by utilizing automated classification and narrative analysis. Our prototype significantly reduces instructor preparation time while improving the diversity, quality, and pedagogical alignment of curated stories. This innovation enables educators to more efficiently source, annotate, and deploy impactful data narratives tailored to their teaching and research objectives.