Story at a glance:
- A graduate student in Switzerland found that Twitter’s cropping feature prefers younger, slimmer and lighter faces.
- The student who revealed the bug won $3,500 in a “bug bounty.”
- The cropping feature even favors white dogs over black dogs.
Twitter’s algorithm favors younger, slimmer and lighter faces over those who have bigger, older and darker features, a student in Switzerland discovered.
Bogdan Kulynych, a graduate student at Switzerland’s EFPL university, revealed the algorithm’s bias during a competition at a security conference in Las Vegas. For his feat, Twitter is paying Kulynych $3,500, The Guardian reported.
The competition at the DEF CON security conference was part of Twitter’s first “algorithmic bug bounty.”
America is changing faster than ever! Add Changing America to your Facebook or Twitter feed to stay on top of the news.
To prove the bias, Kulynych artificially generated faces with different features and then ran them through Twitter’s cropping algorithm to study how the software focused on each image, according to The Guardian. After running the tests, it became clear that Twitter’s algorithm worked against Black faces.
“When we think about biases in our models, it’s not just about the academic or the experimental … but how that also works with the way we think in society,” Rumman Chowdhury, the head of Twitter’s AI ethics team at the conference, told The Guardian.
“I use the phrase ‘life imitating art imitating life’. We create these filters because we think that’s what ‘beautiful’ is, and that ends up training our models and driving these unrealistic notions of what it means to be attractive.”
This is not the first time Twitter’s software has been associated with algorithmic bias. The company had to apologize in 2020 after users noticed that its cropping feature favored white faces over Black faces — even white dogs were treated more favorably in the cropping process than black dogs, The Guardian reported.
This launched the company’s bug bounty.
“Algorithmic harms are not only ‘bugs.’ Crucially, a lot of harmful tech is harmful not because of accidents, unintended mistakes, but rather by design. This comes from maximisation of engagement and, in general, profit externalising the costs to others,” Kulynych said, adding that he had mixed feelings about winning the grand prize. “As an example, amplifying gentrification, driving down wages, spreading clickbait and misinformation are not necessarily due to ‘biased’ algorithms.”
READ MORE STORIES FROM CHANGING AMERICA
THE SIGNS YOU HAVE THE DELTA VARIANT ARE DIFFERENT THAN ORIGINAL COVID-19
EVIDENCE MOUNTS THAT DELTA VARIANT IS DANGEROUS FOR KIDS
THERE’S ONE KIND OF MASK THAT WON’T PROTECT YOU FROM THE DELTA VARIANT
NEW POLL DISCOVERS WHO AMERICANS BLAME FOR CURRENT COVID-19 SURGE
TRUMP SLAMMED BY HIS OWN FORMER HHS SECRETARY OVER PRIVATE VACCINATION
changing america copyright.
