Experiments must be reproducible and lead to the same result, otherwise they are not scientific. What “reproduce” and “same result” actually mean has always been a process of negotiation in the sciences. And the resulting standards have always had a political component.

Computers are not only valued because they are fast, but also because they are able to do “the same thing over and over again” – as is expected of machines. But expectations are disappointed. The use of computers has become so complex and opaque that results are difficult or impossible to reproduce. Are these results then not scientific? Or do researchers need to negotiate new standards? Last but not least: Who has (and who claims) the political power to decide this?

We want to clarify these questions. To this end, we look at how the standards of reproducibility have been negotiated and established at present and in history. We focus on the fields of computational chemistry and thermodynamics, in which such negotiation processes are currently taking place. Also, because something amazing is happening there: Despite the difficulties with reproducibility, the calculated predictions seem to be getting better and better. Is predictive power being placed above scientific rigor here?