AI predicts orientation—a groundbreaking capability demonstrated by Michal Kosinski, a psychologist at Stanford University. Described as the “Oppenheimer of Artificial Intelligence” by Business Insider, Kosinski has shown that AI can uncover hidden, private details such as sexual preferences and political beliefs. With just a few of your Facebook likes and facial images, AI can predict these aspects with surprising accuracy, writes media theorist Ana Martinoli for Radar.

Historical Context and AI Advancements

Throughout history, there have been many attempts to measure someone’s psyche and mind through facial dimensions and bone structure. Kosinski, using AI, continues to develop these theories in practice. He claims that by combining data and facial recognition tools, today’s technology can accurately decipher our attitudes, beliefs, and preferences. Kosinski is aware of the potential misuse of these AI tools. He primarily warns about the consequences in his papers: “Very soon, we could find ourselves in a position where these models have properties and capacities far more advanced than what people can imagine,” he told Business Insider.

Facebook Likes and Personality Predictions

One of Kosinski’s first studies to gain public attention analyzed the behavior and likes of Facebook users. The study showed that computer assessments of individuals’ character and personality, based on their digital footprints, were more accurate than judgments made by close acquaintances or friends. Kosinski’s findings revealed that AI can automatically predict personality and character without relying on human input.

AI and Sexual Orientation Detection

A few years later, Kosinski conducted a study showing that facial features contain more information about sexual orientation than the human brain can detect. Using AI and mathematical systems to analyze over 35,000 photographs, the so-called classifier accurately distinguished between gay and heterosexual men in 81% of cases and between women in 74% of cases.

Predicting Political Orientation with AI

In 2024, Kosinski proved that facial recognition technology could predict a person’s political orientation with astonishing accuracy. His research, published in the journal American Psychologist, shows that even neutral facial expressions can reveal clues about a person’s political beliefs. Kosinski emphasizes that this finding raises significant privacy concerns, especially because facial recognition can operate without an individual’s consent. As facial recognition technologies become more widespread, the risk of using these tools for purposes beyond simple identification, such as predicting political orientation, grows. For more on this topic, check out Kosinski’s full research and related studies on AI and privacy.

“Facial images can be easily and secretly collected by law enforcement agencies or obtained from digital or traditional archives, including social networks, dating platforms, photo-sharing websites, and government databases,” Kosinski stated in an interview with PsyPost. While his research is not conclusive, it highlights an unprecedented threat to citizens’ privacy.