Presentation Information
[B-11-12]A Study on Changes in Identifier Naming Styles Associated with the Growing Use of Large Language Model
◎Yuto Tamura1, Sho Tsugawa1 (1. University of Tsukuba)
Keywords:
Large Language Models,Coding Styles,Software Development
In recent years, the use of large language models (LLMs) in software development has expanded rapidly. While LLMs support individual development tasks, they may homogenize coding styles that have traditionally differed across developer communities by presenting similar code examples to many developers. In particular, identifier naming tends to reflect the conventions of developers and projects, making it an important target for capturing changes in the distribution of coding styles associated with the spread of LLMs. In this study, we analyzed temporal changes in identifier naming styles during the period of LLM adoption, using public projects on GitHub as the target. From projects whose primary languages were Python, JavaScript, TypeScript, Rust, Java, and Go, we extracted variable names, function names, and argument names, and calculated distances between projects’ naming styles based on identifier length, naming conventions, number of words, and related features. As a result, the distances between projects decreased in many languages from around 2024, indicating a trend toward greater similarity in identifier naming styles.