I’m not saying reproducibility is an unworthy aspiration. I merely have issues with the word itself. It’s a misnomer that distracts from the real goals.

You often hear about the need for “reproducible science”, or better yet, a “crisis of reproducibility” in science, particularly in the social sciences. Nobody really cares about whether somebody else can exactly rerun the same published experiment. People just want to know if the finding are true. Let’s rename it the “crisis of significance”.

When I migrated from science to data science, so too did the discourse about reproducibility. Nobody needs to run the exact same analysis as somebody else. Nor do they need to train the exact same machine learning model. No, I simply want to run the same code while tweaking its parameters. Last month’s report needs to be rerun this month. Let me ablate some decisions in somebody else’s model. We don’t need reproducible code, we need reusable code.

Reproducibility feels like a nice to have. The real things we’re after are more important.