Using Generative AI to Create User Stories in the Software Engineering Classroom
Allan Brockenbrough, Dominic Salinas · 2024
A user story is used in agile methodology to describe functionality that is valuable to the user and may include criteria to determine if the developer has completed the story. This study investigates undergraduate computer science students using ChatGPT to create user stories from user feedback. The study compares aspects of the user stories created by students using ChatGPT with those not using ChatGPT. Are user stories written by students with AI assistance of higher or lower quality? How does the time spent writing the user story change with the use of ChatGPT? We evaluate student user stories using a modified INVEST story rating system. Evaluated user story properties include structure, independence, value, testability, and grammar. The results show that ChatGPT-assisted students produce higher-quality user stories than unassisted students. However, using ChatGPT to write user stories does not guarantee high quality. ChatGPT can fail to recognize dependencies between user feedback and create structurally incorrect user stories. We see a need for students to be trained in effectively using this tool by carefully examining AI-assisted output and making revisions.