I mean, replication failures are a result of bad science, bad analysis and a poor understanding of statistics in many cases. It's not so much a crisis in the state of human knowledge but the state of the education of the scientists in those other disciplines.
The phenomena are related, though. One could argue that the social sciences, with greater latitude in definitions of measurement and thresholds for statistical significance, had been building sand castles for decades. Eventually, the entire rotten edifice will collapse.
In contrast, the greater restrictions of theory and measurement in physical science didn’t easily allow researchers to do pointless hand-waving that looked good. In some sense, the “crisis” in physics is less embarassing, as it is simply theorists bumping into limits enforced by reality. They didn’t make it up as they went along.
I’m also doubtful the lauded “open science” movement will accomplish anything besides the mass transfer of intellectual property to centralized data platforms, to be mined by replication specialists.
I’m biased, admittedly, after watching an APA zoom conference on the advantages and wonders of open science, and why researchers should join in. The lead presenter’s #1 reason to join open science was “that I didn’t lose my data anymore, it was all nicely and neatly centralized on the OSF server.”
In biology, neuroscience, and psychology replication failure is often the result of an impedance mismatch between reductive models and real systems complexity.
This is an acute problem in animal research where 90% of work gets away with using a single genotype in a single environment. And then the “wise” old heads at NIH wonder that there is a replication bias? Really?
