Of course the S*n is what it is, and it will look at everything it considers publishing through the lens of “Will this give your everyday fascist OAP man a semi?”
And the scientist helpfully ran their machine learning on the hot celebs, for the edification of s*n readers.
Now, I’m not an expert, but if I know exactly what goes into making the score, I would hesitate to call it machine learning.
AWAT. (By the way did anyone check if the weak increase in “trustworthiness displays” in the original “dataset” could be explained by the occurrence of more women’s portraits?)