Facial recognition tech accurately predicts person's political beliefs: Study
Turns out, facial recognition technology can successfully predict the political stance of people based on their Facebook profile pictures
Think facial recognition has gone too far as a node of edge-cutting science? You may be right, or left, and a new study suggests that facial recognition knows it!
Turns out, facial recognition technology can successfully predict the political stance of people based on their Facebook profile pictures. The software's algorithm exhibited accuracy in terms of a person's political beliefs which remains way higher than a personality test, or just a wild guess.
Over the last few years, scientists have been able to advance facial recognition technology, with the software effectively being able to identify and track individuals.
The study was authored by Michal Kosinski, who is an associate professor at Stanford University. He told PsyPost that he has been studying privacy risks over algorithms over the last decade. Naturally, critics have expressed worries over the privacy safety concerns imbued into facial recognition technology.
The study took into account 1,085,785 faces obtained from social media profiles. Out of these, 977,777 emerged from users in the US, UK, and Canada who had already reported their political beliefs. The other 108,018 faces came from Facebook across the US who also reported their political leaning while completing a personality test with 100 items.
To ascertain the political leanings of users, the algorithm compared the facial features of liberals and conservatives. Turns out, the software was mostly right. Just among Facebook users in the US, the accuracy hit 73 per cent.
According to Kosinski, this result yields better accuracy than humans, who can distinguish between right-leaning and left-leaning people with 55 per cent accuracy.
The study was published in Scientific Reports on January 11, 2021, titled "Facial recognition technology can expose political orientation from naturalistic facial images".