The Perception of ChatGPT and GitHub Copilot in Software Development - A Case Study in a German SME

Kurzfassung

The use of large language models (LLMs) in the context of programming is widely discussed in the scientific literature. However, their use in productive settings (in contrast to for instance teaching) is discussed far less and a focus on the developers' sentiment is even scarcer. To explore how professional programmers perceive the use of those tools and also how this perception changed after utilizing them for some time, a case study within a German SME was conducted that comprised experimental phases (using ChatGPT and GitHub Copilot), corresponding surveys, and a group discussion. Hereby, the developers initially showed a modest positive sentiment towards the use of the tools in the pre-survey that greatly improved within the post-survey, indicating a positive usage experience. Further, the tools' primary use cases, differentiating aspects as well as factors affecting their adoption were determined.