Twitter Is Focusing On More ‘Feminine’ Faces In Photos


Twitter, which caused racism discussions with its photo cropping algorithm, came up with a different problem in the same algorithm. It was announced that Twitter’s biased algorithm focuses on the younger, lighter and feminine faces in the photo.

More CEOs lately Jack Dorsey’nin The social media giant that appears with its statements in the crypto money industry Twitter, has remained a topic of discussion for some time due to algorithm-based problems in its implementation. The algorithm that provides cropping of the pictures is biased and therefore highlighting white-skinned individuals, the social media giant caused a great reaction.

Twitter, which allows photos to be cropped manually but can’t get away with it, racism He started a competition to find out the biases of his algorithm that caused such discussions. Now we are here with the results of the competition.

Younger, brighter and feminine faces are preferred:


Awarded by Twitter Bogan Kulynychtested Twitter’s cropping algorithm with human faces enhanced by artificial intelligence. These developed faces were differentiated from each other by many factors such as skin tone, age and weight. The results obtained from the study in which these different faces were tested, 37 percent lighter skin tone He showed that he prefers photographs with

Besides, the algorithm 25 percent more feminine than 18 percent It was also noted that he preferred young-looking faces.

In the study, the algorithm not examined from all aspects is underlined. In addition, it is not known whether the presence of different backgrounds in the photographs used will give misleading results. Because the work is done by a single person, it is also a question whether the person will include his own thoughts in the work.


Lowering the Dam in the YKS Exam Was Not Welcomed: Here Are the Reactions From Social Media

Twitter will be pleased with the result that Kulynych with exactly 3500 dollars rewarded. If you would like to review Kulynych’s work in detail from here You can reach the GitHub page.

Source :