For example: > As a group, for instance, Latino men rated Latinas as 13 percent more attractive than the average for the site, while they rated African-American women 25 percent less attractive.
That is an insane generalization going on.
What metric are we using to determine this? Is it possible that people who tend to participate in the rating of the looks of potential mates are more inclined to align with race? Not purely Latino men. Which is an enormous initial generalization to be making at the outset.
> Witness the actions of 35-year-old heterosexual men on OkCupid. These men typically search for women between the ages of 24 and 40, Mr. Rudder reports, yet in practice they rarely contact anyone over 29.
Again on this one, was age the only possible metric that caused the under 30 to be contacted more by 35 year old men? There isn't anything else that might be different about an under 30 profile that causes more communication to occur?
This is the actual scary stuff to be publicly releasing as real science. Just as the general public is ill informed about the experiments going on, they are also not aware of what metrics are used to determine these results. In my experience, many of these metrics are not as concrete as the appear and full of pushing the data to fit a narrative etc.
Combine that with the fact that most of this data is proprietary and private with no way to be peer reviewed. Dangerous stuff.