Twitter has recently announced that it would stop using its AI image-cropping algorithm after finding out that the system was biased.
Indeed, following complaints from users that the automated system was having a preference towards pictures of white people, the company decided to test its algorithm. They realized that it had gender-and race-based biases, where the system was automatically favoring white people over Black people and women over men.
The image-cropping system featured a face-detecting software, which was, however, prone to errors. The company decided to change approaches and went on having its software in saliency, meaning that it would detect the most interesting area to look at on a picture. This algorithm was strongly criticized for its biases and Twitter declared it didn’t notice any racial or gender bias during testing before releasing the preview feature.
Yet, it took months before the company decided to properly look into the algorithm. This then showed that not every system on the social media platform can work and that is something better done by humans.
Hence, a few months ago, Twitter started to test a new way to show a full image on mobile devices instead of being cropped, which received positive feedback.
