Ever notice how on some social platforms like Twitter or Instagram that you yourself are mysteriously unable to crop your display images on your own? That’s because Twitter prefers to let their algorithms make the decision. Over the weekend users on Twitter discovered the surprising dangers of letting algorithms crop your own images.
Education tech researcher Colin Madland drew attention to the issue while speaking out about how the video-calling program Zoom, often crops the head out of his black person coworker while on calls.
It didn’t take long for Madland and other users to discover that Twitter’s AIs use discriminatory equations to prioritize certain faces as well. In short, the social platform’s AIs prefer white faces over Black ones.
In response to the discoveries, a Twitter spokesperson acknowledged that the company was looking into the issue “Our team did test for bias before shipping the model and did not find evidence of racial or gender bias in our testing. But it’s clear from these examples that we’ve got more analysis to do. We’re looking into this and will continue to share what we learn and what actions we take,” they stated.
Of course, Madland’s discovery is nothing new. In 2019, test results from the National Institute of Standards and Technology revealed that some of the strongest algorithms online were much more likely to confuse the faces of Black women than those of white women, or Black or white men. “The NIST test challenged algorithms to verify that two photos showed the same face, similar to how a border agent would check passports,” Wired points out. “At sensitivity settings where Idemia’s algorithms falsely matched different white women’s faces at a rate of one in 10,000, it falsely matched black women’s faces about once in 1,000—10 times more frequently. A one in 10,000 false match rate is often used to evaluate facial recognition systems.”
Still, it didn’t take long for users on the platform to ask what other physical preferences Twitter has.
Turns out the AIs prefer Ted Cruz with large anime breasts over a normal-looking Ted Cruz.
The user who tested the image of Cruz, found that Twitter’s algorithm on the back end selected what part of the picture it would showcase in the preview and ultimately chose both images of Cruz with a large anime chest.
It’s nothing new that Twitter has its massive problems.
For a platform that so controls and oversees so much of what we consume and how we now operate, it’s scary to know how Twitter chooses to display people with different skin tones. The round of jokes and Twitter experiments by users has only revived concerns on how “learning” computer algorithms fuel real-world biases like racism and sexism.
Notice any needed corrections? Please email us at firstname.lastname@example.org