Bishop Hill recently posted an interesting link to a paper on exaggerated cause/effect correlations in social neuroscience. Essentially the paper was about the problem of how scientists select what to study and how to study it. It isn’t a new problem, but the scientific world is very slow to grasp this particular nettle, for obvious reasons.
Putting it crudely, scientists who study complex natural phenomena have to study their subject within boundaries and assumptions which are inevitably selected to bring out the phenomenon of interest. But the phenomenon may only be real within those selected boundaries and assumptions. In the real world, it may not be a phenomenon at all.
Putting it even more crudely, scientists who study natural phenomena have powerful reasons to cherry-pick without necessarily being aware of it.
One could debate whether this is scientifically right or wrong, but the interesting point is that it may well be inevitable. To study complex natural phenomena we have to simplify by setting boundaries and making assumptions. If we don’t, then we have difficulty reaching specific conclusions and a corresponding difficulty getting papers published. Why? Because the cherry-pickers are bound to get there first.
And the lesson is? The science of the natural world is mostly about cherry-picking, because the cherry-pickers get there first. Not necessarily through nefarious intent - but they still get there first.