Analysis of how code duplication changed in recent years…
0 reactions
Analysis of how code duplication changed in recent years
My methodology and data set didn't show any trend, but it demonstrated a more important issue: how wrongly this kind of research can be done and how misinterpreted the conclusions can be.
The reason for making this research was an attempt to verify the claim that AI-assisted development increases code duplication. I analyzed 14 well-maintained open-source projects between 2021-2026, excluding new ones developed only with AI. For duplication detection, I compared semantic similarity using https://github.com/rafal-qa/slopo (I'm the author), not exact copies. This data can't prove or deny the claim, no trend is visible. Not only because 14 projects is too little, but also because there is a large variance between projects.
The main advantage of this research is that it highlights the pitfalls in the analysis and conclusions and shows how easy it is to create "evidence" to support any claim.
#technology #dev #programming
source
My methodology and data set didn't show any trend, but it demonstrated a more important issue: how wrongly this kind of research can be done and how misinterpreted the conclusions can be.
The reason for making this research was an attempt to verify the claim that AI-assisted development increases code duplication. I analyzed 14 well-maintained open-source projects between 2021-2026, excluding new ones developed only with AI. For duplication detection, I compared semantic similarity using https://github.com/rafal-qa/slopo (I'm the author), not exact copies. This data can't prove or deny the claim, no trend is visible. Not only because 14 projects is too little, but also because there is a large variance between projects.
The main advantage of this research is that it highlights the pitfalls in the analysis and conclusions and shows how easy it is to create "evidence" to support any claim.
#technology #dev #programming
source
| 2,000 mlx | |
| 5 | |
| 0 |