sourcegraph
April 12, 2024

A study in Denmark found that artificial intelligence algorithms can help predict a person’s political ideology based on their facial features.

The technique found that right-wing politicians were more likely to have happy facial expressions in photos, while people with neutral facial expressions were more likely to be identified as left-wing, the study said.

The study, “Predicting Ideology from Mugshots Using Deep Learning: Expression, Beauty, and Extrafacial Information,” found that AI could predict a person’s political ideology with 61 percent accuracy when analyzing their photos .

In their paper, the researchers explain that deep learning, a method in artificial intelligence in which computer scientists teach computers to learn and process information similar to humans, can be used to make predictions about people based solely on photos, Published in Scientific Reports.

According to the researchers, the scientists sought to determine exactly “what information contributes to the predictive success of these techniques.”

Humans can read another person’s face and make almost instant judgments about personality, intelligence, and even political ideology. Study author Stig Hebbelstrup Rye Rasmussen of Aarhus University and colleagues explored whether computational neural networks — algorithms that mimic the structure and function of the human brain — could predict a person’s political ideology from just a single photo.

AI-driven facial recognition boom raises concerns about bias, privacy: ‘You’re being recognized’

A study in Denmark found that artificial intelligence algorithms can help predict a person’s political ideology based on their facial features. (Smith Collection/Gado/Getty Images)

The scientists trained the neural network using thousands of photos of politicians from the country’s 2017 municipal elections, noting that the elections were neither highly polarized nor contested, and calling the politicians “the last amateurs in politics”.

How to contain the artificial intelligence threat?Let go of the lawyer

They removed any photos of candidates who were not clearly left-wing or right-wing, of non-European ethnic origin or with beards. The photos only depict the candidate’s facial features, with no background photos that could change the predictions. The researchers then left behind 4,647 photos of political candidates, 1,442 of which depicted female politicians.

artificial intelligence logo

Experts believe that the difference between Chinese and U.S. AI investments is that the U.S. model is driven by private companies, while China takes a government approach. (Josep Lago/AFP via Getty Images)

The researchers used Microsoft’s facial expression recognition technology to measure emotional states in photos, along with other algorithms to determine a candidate’s attractiveness and even masculinity. They also used photos of some Danish MPs to test the algorithm’s accuracy.

who is watching youAI can track unsuspecting victims with ‘ease and precision’: expert

Danish general election

People vote during European Parliament elections on May 26, 2019 in Copenhagen, Denmark. (Ida Marie Odegard via Reuters)

All told, the study found that AI trained on data could accurately predict ideology up to 61 percent of the time — suggesting that algorithms can predict political affiliation better than pure chance.

World’s first artificial intelligence university president says technology will upend education and create ‘Renaissance scholars’

“Our results confirm the threat to privacy posed by deep learning methods,” the researchers wrote. “Using a pre-developed, ready-to-use network trained and validated exclusively on publicly available data, we were able to predict the ideology of people in photographs approximately 60% of the time across two samples.”

The study found that more attractive female politicians were more likely to be conservative, while male attractiveness and masculinity were not related to political ideology. The faces of men and women who looked happier were also more likely to be right-wing, while neutral facial expressions meant politicians were more likely to be members of left-wing parties. While more rare, women who show contempt on their faces are more likely to lean left, the study added.

Click here for the Fox News app

“We also demonstrate for the first time that model-predicted ideology is associated with independently classifiable features of the face,” the study said. “For women (though not men), high attractiveness was found among those the model considered likely to be conservative. scores. These results are plausible because previous studies using human raters have also highlighted the link between attractiveness and conservatism.”



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *