AI technology can predict a person’s political views from a single photograph of them, according to a recent study.
In a study published in American Psychologist, researchers developed an AI algorithm that can accurately predict an individual’s political views based solely on a photograph of them.
For the study, almost 600 participants were photographed with the same blank expression and in the same black t-shirt. They were also asked to pull their hair back, remove jewelry and make-up, and shave any facial hair for the photographs.
Nonetheless, AI technology was able to spot tiny nuances in a person’s facial features that correlates to their political learning when they posed expressionlessly in a photograph .
In the study, the researchers found that liberals tended to have smaller lower faces, their chins were smaller and their lips and noses pointed downward, while conservatives have larger, wider features in the lower halves of their faces.
‘People Don’t Realize How Much They Expose in a Photograph’
The researchers say that the findings of the study reveal the potential risks of facial recognition technology to personal privacy.
“I think that people don’t realize how much they expose by simply putting a picture out there,” Dr. Michal Kosinski, the study’s lead author and an associate professor of organizational behavior at Stanford University’s Graduate School of Business, says in an interview.
Kosinski adds: “You can still go to Facebook and see anybody’s picture. This person never met you, they never allowed you to look at a picture, they would never share their political orientation … and yet, Facebook shows you their picture, and what our study shows is that this is essentially to some extent the equivalent to just telling you what their political orientation is.”
Does Your Selfie Say ‘Conservative’ or ‘Liberal’?
For the study, 591 participants filled out a political orientation questionnaire. After the 100-item questionnaire, the participants were photographed wearing a black T-shirt adjusted using binder clips to cover their clothes, removed all jewelry, shaved their facial hair, and wiped off make-up if necessary.
The participants’ hair was also pulled back using hair ties, hair pins, and a headband while taking care to avoid flyaway hairs — all so that the AI software could only focus on their faces.
The “VGGFace2” – an AI facial recognition software created in 2017 – was used to examine the images and determine the participant’s facial characteristics by cross-referencing them with other images stored in a database.
The AI then captured what Kosinki described as a numerical “fingerprint” of the participants’ faces and compared them to a database of their responses to predict their views.
Aside from the projected outlines of conservative versus liberal faces, the study found that all other “outlines and faces are virtually identical, revealing no other obvious differences between liberals and conservatives, including in facial expression, grooming, skin color, or head orientation.”
The findings took into account other factors like the participants’ age, gender, and ethnicity, but researchers found the connection between the person’s political leanings was solely based on their facial characteristics.
Image credits: Header photo licensed via Depositphotos and Center Image via “Facial Recognition Technology and Human Raters Can Predict Political Orientation From Images of Expressionless Faces Even When Controlling for Demographics and Self-Presentation” by M.Kosinki et al. published in American Psychologist