Paper Number
2930
Paper Type
Complete
Abstract
Generative AI is likely to transform processes of innovation, yet it is unclear what trajectory this transformation will follow. The uncertainty arises because certain problems, where a provision of detailed context results in precise solutions, are best approached through interpolative (within-the-box) thinking, whereas other problems, where providing detailed context is either difficult or ineffective, demand extrapolative (out-of-the-box) thinking. How GenAI affects tasks needing extrapolative vs. interpolative thinking is an open empirical question. We study this question within the open-source context, focusing on the launch of GitHub Copilot. Specifically, we leverage a natural experiment in which GitHub Copilot selectively rolled out support for Python but not for R. We observe a significant jump in overall contributions, suggesting that GenAI effectively enhances outcomes for voluntary contributors. Interestingly, Copilot’s launch increased maintenance-related contributions, which need interpolative thinking, significantly more than code-development contributions, which need extrapolative thinking.
Recommended Citation
Yeverechyahu, Doron; Mayya, Raveesh; and Oestreicher-Singer, Gal, "The Impact of Large Language Models on Open-source Innovation: Evidence from GitHub Copilot" (2024). ICIS 2024 Proceedings. 38.
https://aisel.aisnet.org/icis2024/diginnoventren/diginnoventren/38
The Impact of Large Language Models on Open-source Innovation: Evidence from GitHub Copilot
Generative AI is likely to transform processes of innovation, yet it is unclear what trajectory this transformation will follow. The uncertainty arises because certain problems, where a provision of detailed context results in precise solutions, are best approached through interpolative (within-the-box) thinking, whereas other problems, where providing detailed context is either difficult or ineffective, demand extrapolative (out-of-the-box) thinking. How GenAI affects tasks needing extrapolative vs. interpolative thinking is an open empirical question. We study this question within the open-source context, focusing on the launch of GitHub Copilot. Specifically, we leverage a natural experiment in which GitHub Copilot selectively rolled out support for Python but not for R. We observe a significant jump in overall contributions, suggesting that GenAI effectively enhances outcomes for voluntary contributors. Interestingly, Copilot’s launch increased maintenance-related contributions, which need interpolative thinking, significantly more than code-development contributions, which need extrapolative thinking.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.
Comments
14-DigitalInnovation