Thursday 20 March 2025
Software engineering students are often tasked with creating user stories, which describe new features or functionalities of a product from the user’s perspective. These stories must be clear, concise, and testable, making them a crucial part of the software development process.
A team of researchers set out to investigate whether large language models (LLMs) could aid students in crafting better user stories. They recruited 20 undergraduate computer science students and asked them to create user stories for a social media application based on real user feedback. Half of the students were given an LLM, specifically ChatGPT 3.5, to help them build their stories.
The results showed that using an LLM significantly improved the quality of the acceptance testing criteria in the student-created user stories. Acceptance testing is a crucial step in software development where developers ensure that the product meets the required specifications. With the LLM’s assistance, students were able to create more detailed and testable acceptance criteria.
The study also found that LLMs helped students create more valuable user stories. A valuable story must be clear and concise enough for developers to understand its purpose and scope. The LLM’s guidance enabled students to craft better-written stories that provided a clearer understanding of the user’s needs.
However, the researchers noted that using an LLM did not universally improve all aspects of the student-created user stories. In fact, students who didn’t use the LLM performed better when it came to creating stories with the right scope and size for implementation. These findings suggest that while LLMs can be beneficial in certain areas, they may not be suitable for every aspect of software development.
The study highlights the potential benefits of using LLMs in software engineering education. By providing students with tools like ChatGPT, educators can help them develop essential skills for creating high-quality user stories. However, it’s crucial to recognize that LLMs are only a tool and should be used judiciously.
As software development continues to evolve, it’s likely that we’ll see more innovative uses of LLMs in the industry. For now, this study provides valuable insights into how these models can aid students in creating better user stories, ultimately leading to higher-quality software products.
Cite this article: “LLMs Aid Software Engineering Students in Crafting Better User Stories”, The Science Archive, 2025.
Software Engineering, User Stories, Large Language Models, Chatgpt, Acceptance Testing, Quality, Student Learning, Education, Software Development, Artificial Intelligence.
