A study recently published in Scientific Reports found that facial recognition technology can accurately predict a person’s political stance based on their Facebook profile photo. According to the study authored by Michal Kosinski, the algorithm shows greater accuracy in inferring a person’s political orientation than human judgment or a personality test. A common facial recognition algorithm was applied to images of more than a million people, drawn from their posts on the social media platform, to predict their political orientation by comparing their similarity to faces of more liberal and conservative others.
Kosinski, an associate professor at Stanford University, said he had been researching privacy risks to algorithms for the past decade. While speaking with PsyPost, he said that companies collect data and develop algorithms to generate insights, but are reluctant to reveal how accurate such models are. Kosinski added that facial technology can accurately reveal traits ranging from sexual orientation to personality.
READ: Starlink Internet May Not Work Correctly in India. Here’s why
In Kosinski’s study, a face recognition algorithm was applied to nearly 1,085,795 faces. From this dataset, 977,777 came from dating website users in the US, UK and Canada who had indicated their political orientation. The other 108,018 faces came from Facebook users in the USA, who had also informed themselves about their political orientation and had also completed a 100-point personality test.
The algorithm then compared the facial features of each participant to the average facial features of liberals and conservatives. According to the study, the technology used these similarity measures to determine the likelihood that a participant was either a conservative or a liberal. The result showed that the algorithm was able to predict political orientation alarmingly well and with similar accuracy across countries and social media platforms.
READ: Microsoft says Chinese hackers are remotely looting email inboxes
Accuracy over 70%
The study found that this accuracy was 73 percent for Facebook users in the US and 72 percent for users of US dating sites. Among dating site users in the UK and Canada, accuracy reached 70 percent and 71 percent, respectively. The author of the study found that the algorithm performed significantly better than people, who can only distinguish between a liberal and a conservative with an accuracy of 55 percent.
In addition, the study found that political orientation is linked to certain demographics that are easily observed on the face. The researcher even looked at whether certain traits were associated with political orientation, including facial expression, glasses, facial hair, and head posture. Kosinski found that head orientation had a predictive power of 58 percent, with liberals looking more directly into the camera. The study found that emotional expression had a predictive power of 57 percent, with liberals showing more surprise and less disgust. Kosinski said his findings likely underestimate the intelligence of such technology, adding that greater accuracy would likely be seen with higher resolution images, multiple images per person, or an algorithm specifically designed to identify political orientation.
(Image: @ docrussjackson / Twitter)
READ: Google is updating its productivity software to meet the needs of remote workers amid COVID-19
READ: What happened to Orby TV Service? Was Orby TV out of business?
