Meet the man behind the technology that knows you better than your family
Can an algorithm predict your sexual orientation using your facial features only? The research of Michal Kosinski, Assistant Professor at Stanford and PhD at Cambridge, proves it can - with about 90% accuracy.
In 2013, three scholars from Microsoft Research and the University of Cambridge published a staggering paper. As a response to Facebook’s technology for determining users’ personality based on their profile data patented a year before, Kosinski alongside with two colleagues conducted a study designed to measure the accuracy of such a tool. They not only show how much sensitive and personal information can be read into a list of likes on Facebook but also how this knowledge can be used to gain control over social media users. The paper gained recognition in no time, becoming the second most widely discussed paper in the press and social media in 2013 (Altmetrics.com) and fifth most cited paper ever published by PNAS (Google Scholar). It also made major social media platforms tighten their privacy policies in order to protect the users.
You are what you like
Conducted on over 58,000 volunteers, who shared information on their likes, detailed demographic profiles, and the results of several psychometric tests, the study proved 88% accurate for determining male sexuality, 95% accurate distinguishing African-American from Caucasian American and 85% accurate differentiating Republican from Democrat. Christians and Muslims were correctly classified in 82% of cases, and good prediction accuracy was achieved for relationship status and substance abuse – between 65 and 73%. These revelations however, seem to be only the tip of the iceberg.
With striking efficiency the scholars managed to grasp personality traits including intelligence, emotional stability, openness and extraversion. Liking curly fries on Facebook and high IQ? Apparently the two can be linked by the algorithm which, as the authors of the study observed, may reveal personal information with more exactness than one’s real life family and friends.
The groundbreaking research put Kosinski in the limelight. He became a target of morally dubious propositions from politicians and others wanting to use his algorithms to gain commercial influence or political power over social media users. His intention though was to merely show that any internet activity - from Facebook likes through music playlists, dating profiles, browsing history and credit card records - conveys enormous amounts of insight into our minds. As Kosinski claims: “I did not build the bomb. I only showed that it exists.”
Your face looks gay
The scientist didn’t stop there. Having established a reputation of an innovator, he coauthored another controversial study, this time with the results of running 35,326 facial images through an algorithm. It revealed surprising links between facial features and sexual orientation. “Given a single facial image, a classifier could correctly distinguish between gay and heterosexual men in 81% of cases, and in 71% of cases for women. Human judges achieved much lower accuracy: 61% for men and 54% for women”.
The results of the study made it to the cover of The Economist, caused a turmoil within the LGBT community, resulting in a statement from Human Rights Campaign and Glaad (formerly known as the Gay and Lesbian Alliance Against Defamation) and opened a discussion on how the A.I. knows what it knows. Even Kosinski himself said that “he was shocked it was so easy” for the algorithm to denote such conclusions from the data.
The Guardian reports that Kosinski’s friends describe him as “a brilliant, provocative and irrepressible data scientist who has an insatiable (some say naive) desire to push the boundaries of his research”. Probably, a few of his Facebook likes with the help of his own algorithm would tell us more about his personality, but about his work and the privacy in the digital world it is most worth to listen to him talk for himself.
Written by Agnieszka Dzierżak, Dare Magazine