Left and right distinction is now possible via facial recognition

Researchers developed an AI algorithm that in 70% of the cases on a face can distinguish left and right in voting behavior.

The election battle between left and right in USA ran high at the end of 2020. Source: Wikipedia

In discussions between left and right, it often seems as if two different human species are talking to each other. Nowadays left and right talk more and more past each other.

They see the other as alien. Mostly left has a hand in this and reacts more emotionally. An important distinction is also that left-wing voters show more emotions in their face than right-wing voters.

Not clear why left and right look different

The annoying thing about neural networks is that they can make a reasonably accurate prediction, but not on the basis of which they make that prediction. The researchers were therefore unable to pinpoint exactly what kind of facial features their system correlated with political affiliation, but they did find some trends. Like the attitude of someone's head.

The degree to which some form of facial expression could be seen in the photos also seemed to provide some clues. For example, people who looked directly into the camera, a favorite hobby of ex-VVD politician Wiegel, who is known for his conservatism, usually judged the algorithm to be more progressive. Those who looked away and showed disgust, therefore ended up on the “conservative” pile.

Difficult to replicate

Repeating this research by others, an absolute requirement in science, is difficult. After all, the photos and political preferences on which their algorithm has been trained contain pretty privacy-sensitive information. Other researchers will therefore have to create their own collection of left and right voters and apply the algorithm to this.

That is only going to be difficult, now that the proportions are so polarized. And this correlation may not be correct outside the (California) target groups from which the researchers obtained their data ...

How can you use this knowledge as a spin doctor?

The intention is to allow voters of your political preference to vote, and to discourage others from voting. For example, you could identify left and right voters by means of facial recognition. If you want mainly left-wing voters to vote, you can appeal to them and encourage them to vote. You want to deprive right-wing voters of the desire to vote with cynical jokes about the “pocket-filling politicians”.

Michal Kosinski. Facial recognition technology can expose political orientation from naturalistic facial images, Scientific Reports (2021). DOI: 10.1038 / s41598-020-79310-1

Leave a Comment