The recent conversation around the Twitter image cropping methods led to reviewing the test for bias in the systems and improving how images are displayed. Twitter has announced the development of a solution for these areas.
Twitter tested the existing machine learning (ML) system and image cropping methods that decide how to crop images before bringing it to Twitter.
A previous experiment of the image-cropping method on the platform:
Twitter says the image cropping system relies on saliency, which predicts where people might look first. For the initial bias analysis, they tested pairwise preference between two demographic groups (White-Black, White-Indian, White-Asian, and male-female).
The platform claims their analysis to date hasn’t shown racial or gender bias, but they recognize the way they automatically crop photos means there is a potential for harm.
They are currently conducting additional analyses to add further rigor to the testing.
Twitter intends to decrease the reliance on ML-based image cropping by giving people more visibility and control over what their images will look like in a Tweet.
They’ve started exploring different options to see what will work best across the wide range of images people Tweet every day. Giving people more choices for image cropping and previewing what they’ll look like in the Tweet composer may help reduce the risk of harm.
The photo users see in the Tweet composer is what it will look like in the Tweet. There may be some exceptions to this, such as photos that aren’t a standard size or are really long or wide. In those cases, there is a need to experiment with how the photo is presented in a way that doesn’t lose the creator’s intended focal point or take away from the integrity of the photo.