Examining Software Engineering Practices in the Pre-AI and Post-AI ERA

Abstract

The rise of generative artificial intelligence (AI) coding tools such as GitHub Copilot and ChatGPT has reshaped software development, yet their impact on open-source software quality remains unclear. This study conducts a longitudinal analysis of code review and bug-fix patterns across six major Python and JavaScript repositories, pandas, scikit-learn, TensorFlow, Django, React, and Node.js, comparing the pre-AI (2018-2021) and post-AI (2022-2025) periods. Using GitHub pull request data, we examine changes in reviewer participation, review duration, and comment density, alongside post-merge bug-fix frequencies. Results show a mild decline in review intensity after widespread AI adoption, alongside stagnant or increased bug-fix activity, suggesting no corresponding improvement in software quality. These findings suggest potential overreliance on AI-generated code and highlight critical trade-offs between development speed and code robustness, offering empirical evidence for a more cautious, evidence-based integration of generative AI tools in software engineering practice.

Ähnliche Arbeiten