Edit-I see people may take this as a joke. Sorry, not funny. Especially not right now.
> "One of my favorite¹ concepts for multi-dimensional data visualization is the Chernoff Face"
> 1: "Favorite" might be code for "useless," going with the theme of this blog"
Don't act like this project will have actual real world uses and ramifications. No, it won't literally teach people to associate good or bad qualities with facial features, because nobody will ever use it for practical purposes. No, there's no liability to promote prejudice, this is just some programmers ML side-project.
> How is that not objectifying?
It literally is. The generated faces are mathematical objects. They're not real people.
Right now what? Sounds like you're jumping at shadows. In general false positives are good for finding errors in systems. However, they can go too far and lead to pathological patterns like human immune system self-alergies.