AI capable of determining a person’s political affiliation based on their photo, with finds liberals looking at the camera while conservatives have a disgusting look
- Stanford experts have built an AI that can guess political beliefs from a photo
- It was trained with over a million images from dating sites and Facebook
- The AI concentrated on head orientation and facial expressions when guessing
- As it turns out, most liberals are looking at the camera, while conservatives are disgusted
The Stanford study that made headlines in 2017 for designing an AI that uses ‘facial landmarks’ to determine a person’s sexual orientation is back with another potentially controversial system.
Dr. Michal Kosinski claims to have a facial recognition algorithm that can identify whether a person is liberal or conservative based on a single photo – and with over 70 percent accuracy.
The technology, which builds on the AI of 2017, was trained with more than a million images from dating websites and Facebook and programmed to focus on expressions and attitude.
Although Kosinski and his team were unable to determine the exact characteristics of the algorithm related to a political bias, they did find some trends such as head orientation and emotional expression in photos.
Some examples include that people who looked directly into the camera were labeled as liberal and those who showed disgust were viewed as more conservative.
Scroll down for video
The technology was trained with more than a million images from dating websites and Facebook and programmed to focus on expressions and attitude. The machine learning system crops and resizes the face to reduce capturing of non-facial features
The study, published in Nature, states that when people are asked to distinguish between two faces – one conservative and one liberal – they are right about 55 percent of the time.
“Because people may miss or misinterpret some clues, their low accuracy does not necessarily represent the limit of what algorithms could achieve,” the study read.
Algorithms excel at recognizing patterns in huge datasets that no human could ever process, and outperforming us at visual tasks ranging from diagnosing skin cancer to facial recognition to face-based assessments of intimate characteristics, such as sexual sexual orientation (76% vs. 56%) 7, personality (64% vs. 57%; derived from Pearson’s rs), and – as shown here – political orientation. ‘
Researchers used a sample of 1,085,795 participants from the US, Canada, and the UK, along with their self-reported political orientation, age, and gender.
The Stanford study that made headlines in 2017 for designing an AI that uses ‘facial landmarks’ to determine a person’s sexual orientation (pictured) is back with what may be another controversial system.
The study notes that the ethnic diversity of the same included more than 347,000 non-white participants.
The machine learning system crops and resizes the face to reduce the capture of non-facial features.
When identifying US images, the AI was 72 percent accurate.
Similar accuracy was seen in the Canadian sample, 71 percent, and the UK, at 70 percent.
Researchers used a sample of 1,085,795 participants from the US, Canada, and the UK, along with their self-reported political orientation, age, and gender. When identifying US images, the AI was 72% accurate. A similar accuracy was seen in the Canadian sample, 71%, and the UK with 70%
The highest predictive power was provided by head orientation (58 percent), followed by emotional expression (57 percent).
Liberals tended to look more directly at the camera, were more likely to express surprise, and less likely to express their disgust – those with disgusted gaze were labeled conservative.
In other words, a single facial image reveals more about a person’s political orientation than their answers to a fairly lengthy personality questionnaire, including many items ostensibly related to political orientation (e.g., ‘I treat all people equally’ or ‘I believe that a lot tax money goes to support artists ‘)’, the study reads.