Twitter’s photo cropping algorithm has an extension Biased behavior The company said Wednesday (19), which excludes blacks and men from the photos, adding that the finding indicates that cropping the images is a better decision for people to make.
The study, conducted by three company-based machine learning researchers, was conducted after criticism last year for previewing images published in posts that excluded black faces.
The survey found a difference of 8% in demographic parity in favor of women and 4% in favor of white individuals.
The poll cited several possible causes, including issues related to background images and eye color, but said that none of them is an excuse for the problem.
The authors write: “Image cropping on the basis of machine learning technology is fundamentally flawed because it restricts the expression of the user’s identity and values, rather than imposing a standard view on which part of the image could be considered most interesting.”
To solve the problem, Twitter recently started featuring images in full record proportions – without any cuts – in its mobile apps and is trying to expand this initiative.
The researchers also assessed whether the scraps in the photos favored a woman’s body over their faces, reflecting what is known as a “masculine appearance,” but concluded that this is not the case.
