Konrad HinsenNew blog post: "Going for robustness: science".
How can we make scientific findings and scientific processes more robust?
https://blog.khinsen.net/posts/2025/03/25/robustness-in-science.html
"... it is important not to overlook the most important aspect of the [replicability] crisis: an overestimation of how replicable published scientific work can be expected to be."
https://blog.khinsen.net/posts/2025/03/25/robustness-in-science.html
Bastian Greshake Tzovaras@khinsen oh, I love that point. I’ve been arguing a similar point: if we agree that science should be falsifiable (and many scientists seem to not even know the concept tbh), then we shouldn’t be surprised that we “fail to reproduce” things, but in many cases that just means we’re collectively learning!
@gedankenstuecke Exactly!
That's the argument in principle. The next step is estimating the expected failure rate. That requires a good understanding of one's methods and their reliability, which many disciplines don't have. And that again is fine as long as everyone is aware of it.
@khinsen exactly that! And computational methods are so poorly understood in those “newer” disciplines!