ISR最新发表!生成式人工智能对开源软件协作开发的影响:来自 GitHub Copilot 的实证研究

The Impact of Generative AI on Collaborative Open-Source Software Development: Evidence from GitHub Copilot

Fangchen Song, The University of Texas at Austin

Ashish Agarwal, The University of Texas at Austin

Wen Wen, The University of Texas at Austin

Problem:

How does generative AI change the way people work together to build open-source software? Specifically, we examine how GitHub Copilot helps developers contribute more code and creates new coordination challenges, and whether it benefits some developers more than others.

Relevance: 

Generative AI is rapidly transforming software development. Understanding its effects on coding productivity, team coordination, and how benefits are distributed across developers is important for researchers studying the future of work and for organizations and open-source communities seeking to adopt AI effectively.

Theoretical foundations: 

Our research draws on the cost–benefit and expectancy value model to explain how generative AI affects developers' decisions to contribute code. Moreover, we use coordination theory to explain how AI changes the coordination required to integrate code contributions.

Methodology: 

We conduct a large-scale empirical analysis that combines GitHub’s proprietary GitHub Copilot usage data with publicly available data from GitHub on open-source software projects. We exploit the staggered adoption of GitHub Copilot and estimate its effects using the generalized synthetic control method to identify Copilot’s impact on open-source software development. 

Results: 

Generative AI increases code contributions by encouraging more developers to participate and helping developers contribute more. However, these gains come with increased coordination time. Moreover, the benefits are unevenly distributed, with core developers gaining more from AI than peripheral developers, highlighting the need for organizations and open-source communities to strengthen coordination and better support for less central contributors.

Contributions: 

For scholars, our study shows that generative AI reshapes not only code production but also coordination and the distribution of benefits across developers in collaborative software development. For practitioners, effective adoption requires complementing AI tools with governance and coordination processes that support code integration and less central developers. For policymakers and open-source leaders, our findings highlight the importance of preserving broad participation and ensuring that AI benefits are shared across contributors.

Keywords: 

Generative AI, Open-source Software Development, AI Pair Programmer, Project-level Code Contributions, Coordination Time, Core Developers, Peripheral Developers