Researchers at Arhus University in Denmark have conducted a groundbreaking study suggesting that artificial intelligence (AI) algorithms can predict a person’s political leaning based on their facial characteristics.
The results indicate a 61% accuracy rate in discerning an individual’s political ideology from a single photograph.
The study, titled “Using deep learning to predict ideology from facial photographs: expressions, beauty, and extra-facial information,” was published in Scientific Reports.
It provides valuable insight into the potential of deep learning, a subset of AI that mimics the human brain’s learning process, to make informed predictions from facial analysis alone.
The research team, led by Stig Hebbelstrup Rye Rasmussen, utilized computational neural networks, akin to the human brain’s structure and function, to predict political leanings.
The researchers applied these methods to a collection of photos featuring politicians from Denmark’s 2017 municipal elections, a contest they characterized as the “last amateurs in politics.”
The scientists only included photos of candidates who were explicitly identified as left or right-wing, of European ethnicity, and without beards, to ensure the accuracy of the data.
They focused solely on facial features, excluding any images with distracting backgrounds. Ultimately, the dataset comprised 4,647 photographs, 1,442 of which were of female politicians.
The researchers employed facial expression recognition technology from Microsoft and other algorithms to assess the candidates’ emotional state, attractiveness, and perceived masculinity.
They also tested the algorithm’s performance using a selection of images of Danish parliamentarians.
According to the findings, the AI model trained with these images could predict political affiliations with a 61% accuracy rate, performing significantly better than random guessing.
The researchers acknowledged, “Our results confirmed the threat to privacy posed by deep learning approaches.”
They also noted that “Using a pre-developed and readily available network that was trained and validated exclusively on publicly available data, we were able to predict the ideology of the pictured person roughly 60% of the time in two samples.”
Additional discoveries from the study revealed that right-wing politicians often appeared happier in photos, while left-wing individuals generally had more neutral expressions.
For female politicians, higher attractiveness was linked to conservatism, whereas for men, attractiveness and masculinity were not tied to a specific political ideology.
The study further stated, “For females (though not males), high attractiveness scores were found among those the model identified as likely to be conservative.
These results are credible given that previous research using human raters has also highlighted a link between attractiveness and conservatism.”
More unusually, left-leaning women were found to be more likely to display facial expressions of contempt.
The researchers said, “We also provide the first demonstration that model-predicted ideology connects to independently classifiable features of the face.”
The team hopes that their research will help to understand “what information contributes to the predictive success of these techniques.”
