Twitter confirms its image-cropping AI favours white faces, women

The social media firm added that cropping an image is a decision best made to people.

According to details, Last Fall, Canadian student Colin Madland noticed that Twitter’s automatic cropping algorithm continually selected his face—not his darker-skinned colleague’s—from photos of the pair to display in tweets.

The episode ignited accusations of bias as a flurry of Twitter users published elongated photos to see whether the AI would choose the face of a white person over a Black person or if it focused on women’s chests over their faces.

At the time, a Twitter spokesperson said assessments of the algorithm before it went live in 2018 found no evidence of race or gender bias. Now, the largest analysis of the AI to date has found the opposite: that Twitter’s algorithm favours white people over Black people.

That assessment also found that the AI for predicting the most interesting part of a photo does not focus on women’s bodies over women’s faces.

The microblogging platform was called out in October after its image-cropping model failed to serve all users equitably. Several users noted the algorithm favoured White people and women in photos.

Some instances involved the algorithm choosing a woman’s breast or legs as a salient feature in photos, thereby objectifying women.

The company launched an initiative in April to analyse how algorithms behave on its platform by consulting with engineers, researchers and data scientists.

Twitter in a statement confirmed the biases and said it has decided to abandon using the algorithm to crop images. “One of our conclusions is that not everything on Twitter is a good candidate for an algorithm, and in this case, how to crop an image is a decision best made by people,”

The social network introduced a new way to display standard aspect ratio photos in both Android and iOS devices in March last year.

The feature will allow users to preview an image before tweeting, reducing the dependency on machine learning.