[robbedpeter]

Counterpoint: valid statistical methods and good experimental design has produced good psychological science. Most notably, the predictive power of various theories has been demonstrated in the various social media and advertising algorithms that target different categories of people.

Most of psychology is borderline religious bunkum, but not all. It will eventually converge with a theory of biological intelligence and neuroscience, and until then, will likely not improve much from broad statistical tools and observations.

[necovek]

The understated point is "valid" in "valid statistical methods".

Frequently, social sciences (and even a lot of medicine: basically, anything involving humans where applying valid statistical methods can quickly become unethical) rely on too small a sample set to eradicate the noise.

You do not have to model all factors, because it is impossible to do so, you just need to have a large enough sample to ensure statistical significance that's not out of "statistics in a week" course.

I've seen one too many 50-subjects study, that claims to have extremely high confidence.

[throwawaysea]

> Most notably, the predictive power of various theories has been demonstrated in the various social media and advertising algorithms that target different categories of people.

Can you be more specific? I’m not sure what the field of psychology has to do with algorithms that predict relevance and engagement based on their own data on clicks and other user actions.