I was thinking about the recent “Reproducibility Project” and how it humbly revealed that science is hard, and we really aren’t certain about much at all, let alone having confidence in a single result. Then I started to wonder, are there other domains that, one might say, “got this reproducibility thing down, pat?” The answer, of course, is yes. Now, I’m extremely skeptical of this domain, but the first one that comes to mind is art. Can we learn any lessons for science from art? The metaphor is clear: It was technology that gave power to reproducible art. We can imagine a single statistical result as an effort to capture a scene. A master painter, if for a second you might superficially say his effort was to reproduce such a scene, went out of work with the advent of the photograph. A photographer that develops film was replaced by the printer, and then digital photography exploded that the quantity to insurmountable digits. This is not to say that we should change our expectations for the quality of the photograph. Any kind of result is still valued based on the statistical decisions and experimental design, just as a single picture can be valued based on the quality of the production medium and decisions for the layout. However, the key difference is figuring out how to produce such a result with the quality that we desire en-masse. And just like the photograph, the answer is in the technology. If we want reproducible science, we need better infrastructure, and tools, whether that comes down to the processing of the data or dissemination of the results. If we want reproducibility the responsibility cannot be in the hands of a single artist, or a single scientist. We must do better than that, or things will largely stay the same: slow progress and limited learning. I cannot say that it is not valuable to ask good biological questions, but this kind of thing is what makes me much more passionate about building tools than pondering such questions.
Suggested Citation:
Sochat, Vanessa. "Reproducibility: Science and Art." @vsoch (blog), 31 Aug 2015, https://vsoch.github.io/2015/reproducibility-science-and-art/ (accessed 28 Nov 24).