You're applying the math brain wrong by using the "single counterexample invalidates whole article" mode, rather than just inserting the words "most" or "on average" or "in general" where necessary.

A specific kid will have individual behaviors. A group of kids will have behaviors that can be averaged. Different samples will have different outcomes.

I know sociology has poor reproducibility, but cultural and behavioral differences are definitely a thing.

I used to have a Korean colleague who'd moved to the UK specifically because he did not want his kids growing up in the Korean school system. They will always be ethnically and "genetically" Korean, and I would assume he would teach them the language, but he wanted them to be less culturally Korean because he thought they would be happier that way.