Twitter is under fire by users who claims that the image cropping feature of the app which the company said is meant to highlight ‘salient’ areas of an image predicting where someone is ‘likely to look’ is working against anything black and in favour of white.
Many users posted their experiment with images of black and white people even that of dogs, thereby displaying the obvious difference in the way the cropping feature which is using neural network to make its prediction works, it largedly shows that it centers more on a white face than on a black or coloured face.
Twitter in its defense has written in a blog saying that it has conducted its own experiment and found nothing wrong with feature “our analyses to date haven’t shown racial or gender bias,” though “we recognize that the way we automatically crop photos means there is a potential for harm.” The company also noted that it plans to do additional testing and is also “exploring ways to open-source our analyses so that others can help keep us accountable.”
Trying a horrible experiment…— Tony “Abolish (Pol)ICE” Arcieri 🦀 (@bascule) September 19, 2020
Which will the Twitter algorithm pick: Mitch McConnell or Barack Obama? pic.twitter.com/bR1GRyCkia