But you have less obvious non-representative means: average # of sex partners (Pareto distribution), average life expectancy (bathtub curve), average autism-spectrum quotient (camel curve). In all these cases the averages obscure, rather that elucidate.
I mean if you intentionally increase the entropy of your model, sure. But knowing the average person has 1 testicle and 1 ovary is still better than if you really didn't know anything about that before.
That one is obvious.
But you have less obvious non-representative means: average # of sex partners (Pareto distribution), average life expectancy (bathtub curve), average autism-spectrum quotient (camel curve). In all these cases the averages obscure, rather that elucidate.