Building the DevGPT Dataset for Developer–ChatGPT Studies

This study examines how the DevGPT dataset was created, cleaned, and prepared for research into developer–ChatGPT interactions. Drawing from over 16,000 shared GitHub conversations, researchers filtered out duplicates, non-English exchanges, and limited analyses to eight-turn dialogues. The final dataset offers a rich foundation for exploring how developers use ChatGPT within real-world coding workflows.