Stanford University researchers say that artificial intelligence can deduce sexual orientation by analyzing subtle differences in face structure, The Economist is reporting.
The scientists, who plan to publish their findings in the Journal of Personality and Social Psychology, reportedly downloaded more than 130,000 pictures of men and women from an American online dating site to use as part of the experiment. The profile pictures were public, as were the users’ sexual orientations. They then utilized a combination of facial-detection technology and a predictive statistical model to teach the machine to determine which faces belonged to gay and straight people.
The machine was able to distinguish a gay man from a straight man 81 percent of the time, while that number dipped to 71 percent accuracy for gay and straight women. Using the same photos, regular people could only distinguish men 61 percent of the time, and 54 percent of the time for women.
The researchers reportedly say that the goal of the experiment was not to actually create a software that can “reliably determine” sexual orientation, but to show that it can be done.
This article was featured in the InsideHook newsletter. Sign up now.
