AI able to determine a person’s political affiliation based on their photo, with findings of liberals looking at the camera while conservatives have an expression of disgust
- Stanford experts have built an AI capable of guessing political affiliation through a photo
- Has been trained with over a million images from dating sites and Facebook
- AI focused on head orientation and facial expressions when guessing
- He found that most liberals look at the camera, while conservatives look disgusted.
The Stanford research that made headlines in 2017 for designing AI that uses ‘facial landmarks’ to determine a person’s sexual preference is back with what could be another controversial system.
Dr. Michal Kosinski claims to have a facial recognition algorithm capable of identifying whether a person is liberal or conservative based on a single photo – and with more than 70% accuracy.
The technology, which is based on the 2017 AI, has been trained with over a million images from dating sites and Facebook and programmed to focus on expressions and posture.
Although Kosinski and his team were unable to define the exact characteristics, the algorithm associated it with a political preference, but they found some trends, such as head orientation and emotional expression in photos.
Some examples include people who looked directly at the camera were labeled liberal and those who showed disgust were judged to be more conservative.
Scroll down to video
The technology has been trained with over a million images from dating sites and Facebook and programmed to focus on expressions and posture. The machine learning system cuts and resizes the face to reduce the capture of non-facial features
The study, published in Nature, says that when humans are asked to distinguish between two faces – one conservative and the other liberal – they are correct about 55% of the time.
“As humans may be missing or misinterpreting some of the clues, their low accuracy does not necessarily represent the limit of what the algorithms can achieve”, says the study
‘Algorithms excel at pattern recognition in huge data sets that no human being could process, and are increasingly surpassing us in visual tasks ranging from skin cancer diagnosis to facial recognition to face-based judgments of intimate attributes , such as sexual orientation (76% vs. 56%) 7, personality (64% vs. 57%; derived from Pearson’s rs) and – as shown here – political orientation. ‘
The researchers used a sample of 1,085,795 participants from the United States, Canada and the United Kingdom, along with their self-reported political orientation, age and sex.
The Stanford research that made headlines in 2017 for designing AI that uses ‘facial landmarks’ to determine a person’s sexual preference (pictured) is back with what could be another controversial system
The study notes that its ethnic diversity included over 347,000 non-white participants.
The machine learning system cuts and resizes the face to reduce the capture of non-facial features.
When it came to identifying US images, the AI was 72% accurate.
A similar accuracy was observed in the sample from Canada, 71%, and in the United Kingdom, with 70%.
The researchers used a sample of 1,085,795 participants from the United States, Canada and the United Kingdom, along with their self-reported political orientation, age and sex. When it came to identifying US images, AI was 72% accurate. Similar accuracy was observed in the Canadian sample, 71%, and in the UK, 70%
The greatest predictive power was provided by head orientation (58 percent), followed by emotional expression (57 percent).
Liberals tended to face the camera more directly, were more likely to express surprise and less likely to express disgust – those with a look of disgust were labeled as conservative.
‘In other words, a single facial image reveals more about a person’s political orientation than their responses to a rather long personality questionnaire, including many items apparently related to political orientation (for example,’ I treat all people equally ‘) or ‘I believe that too much tax money goes to support artists’),’ says the study.