Stanford researcher Michal Kosinski has published a new study in which he claims that he can determine the political position of a person with a higher than average probability. Michal is a renowned analyst who has already written several similar studies. One of them was associated with determining a person’s sexual preferences by the face, he also worked on Facebook, in which he helped to compose a psychological portrait of the user by likes.
For a new study, he took millions of faces from open social networks, on many of them users write their political views, so the researcher had a large database. He loaded all the data into a facial recognition algorithm that “broke” each face into 2,000 important metrics with basic features. They are all tied to a conservative or liberal user position.
Michal claims that in 72% of cases, the algorithm correctly indicated a person’s political views. There were many indicators: from glasses to head tilt. For example, liberals are more likely to look directly into the camera and feign surprise, while conservatives are usually aged, fair-skinned, and show disgust on their faces.
The whole thesis of Kosinsky’s research is based on the idea of the science of physiognomy, which some do not consider reliable. She claims that you can recognize the type of person by his face. Psychologists have been saying for years that algorithms that claim to classify whether someone is more likely a bank robber, political scientist, or stubborn Republican based on their face are actually little better than random guesswork.
Michal tries to prove the opposite with his research.