Behold the bizarre return of phrenology.
The research found that gay men and women tended to have “gender-atypical” features, expressions and “grooming styles”, essentially meaning gay men appeared more feminine and vice versa. The data also identified certain trends, including that gay men had narrower jaws, longer noses and larger foreheads than straight men, and that gay women had larger jaws and smaller foreheads compared to straight women.
Human judges performed much worse than the algorithm, accurately identifying orientation only 61% of the time for men and 54% for women. When the software reviewed five images per person, it was even more successful – 91% of the time with men and 83% with women. Broadly, that means “faces contain much more information about sexual orientation than can be perceived and interpreted by the human brain”, the authors wrote.
It’s easy to imagine spouses using the technology on partners they suspect are closeted, or teenagers using the algorithm on themselves or their peers. More frighteningly, governments that continue to prosecute LGBT people could hypothetically use the technology to out and target populations. That means building this kind of software and publicizing it is itself controversial given concerns that it could encourage harmful applications.
https://amp.theguardian.com/technology/2017/sep/07/new-artificial-intelligence-can-tell-whether-youre-gay-or-straight-from-a-photograph
Yeah, loads of uneasy misgivings with this :-/
ReplyDeleteI've had to deal with misfiring gaydar my whole life.
ReplyDeleteI'm certain that there are people I meet IRL regularly who still think I'm closeted or some such.
Daniel Taylor nah, must be the bird, just get rid of it ;-)
ReplyDeleteI take issue with the statement "...than can be perceived and interpreted by the human brain." Humans don't generally get a training set: many gays, even if not actually closeted, don't go out of their way to advertise it, and discussing others' orientation at work is legally risky behavior. The human eye is very good at detecting subtle features (e.g., "sibling-ness"), and I suspect it could do equally well with similar training. (Not necessarily photographs, though; humans pick up a lot from the 3D-ness of direct interaction.)
ReplyDeleteIt's not clear from the article how much is face shape and how much is grooming - where cultural differences between gay and straight people might be expected.
ReplyDeleteThe linked Economist article has far more detail, including that the 91% success rate is only when given an equal sample of gay and straight men and that given s sample with a distribution similar to the general population, less than half of those it picked out as gay actually were, although its most confident picks were back up to 90% but meant it missed a lot of gay men. I.e., as you might expect, you can chose between completeness and reliability.