Twitter is fixing its automatic image Cropping
The company said it has not found evidence of prejudice, but there is “potential harm.” After users complained about its bias, Twitter will change its automatic image cropping feature.
The company said it is working to “reduce reliance on automatic cropping” so that users can ultimately control how their images are displayed in tweets.
Twitter wrote: “We hope that giving people more options for cropping images and previewing the appearance of Tweet composers may help reduce the risk of injury,” Twitter wrote. “Looking forward, we are committed to following the “what you see is what you get” design principle.
Trying a horrible experiment…
Which will the Twitter algorithm pick: Mitch McConnell or Barack Obama? pic.twitter.com/bR1GRyCkia
— Tony “Abolish (Pol)ICE” Arcieri 🦀 (@bascule) September 19, 2020
The meaning is simple: the photos you see in Tweet composers are what they look like in Tweet.” It is not clear when these changes can be implemented.
Twitter first introduced image cropping in 2018, stating that it intended to highlight “significant” areas of an image by using neural networks to predict where someone “may look like”.
However, recently, after some users have made a viral tweet and pointed out that, regardless of the composition of the original image, the crop will always pick out people with lighter skin.
Other users soon used their “experimental” version, which included images of cartoon characters and even black and white dogs.
Twitter said in its blog that it has conducted its tests, “So far, our analysis has not shown racial or gender bias,” although “we recognize that the way photos are automatically cropped means potential harm.”
The company also stated that it plans to conduct more tests and is “exploring ways to open up our analytical resources so that others can help us take responsibility.”