DN-Hypo-Pipeline: An AI-Driven Workflow for Generating Hypotheses using Large Language Models and Scientific Explanations

DN-Hypo-Pipeline is an AI-driven workflow that utilizes large language models to facilitate hypothesis generation by integrating scientific explanations as prior knowledge. It has been evaluated using three highly cited papers in data science, demonstrating superior effectiveness over traditional direct generation methods through both statistical inference and expert evaluation. This pipeline not only aids in deriving novel hypotheses but also provides a theoretical framework that could be extended to various scientific domains, enhancing the modeling process across disciplines.