Twitter is deprecating its photo cropping algorithm in preference of enlarged displays, and according the platform, it is kind of unfair hence investigation is required. The Cropping algorithms were found to be slightly skewed toward particular ethnicities and genders, however Twitter's major complaint appears to be that the AI does not allow users flexibility.
Beginning in 2018, Twitter began cropping photographs to accommodate into its picture preview boxes using the algorithm. The algorithm struggled to find out what a user will find aesthetically appealing so it may incorporate it in the preview. However, some people criticized it for discriminating white faces over black ones. To verify if that is actually the case, Twitter asked some of its researchers conduct a study assessing its crop selection decisions.
On testing the algorithm for gender profiling, researchers discovered that it preferred white women to black women 7 percent of the time, and white males over black males 2 percent of the time. Overall, it has favored white people over black people 4 percent of the time.
Twitter additionally looked for gender discrimination, seeing if the algorithm crop photographs of women wrongly i.e. cropping in an inappropriate way around chest or legs rather than face. It discovered that though the algorithm prioritized women 8% of the time but the cropping is not done inaccurately. In roughly 3 percent of photos where it did not crop the face, other items were focused such as jersey number of sport shirt.
Twitter acknowledges that the algorithms may be damaging in aspects it may not have explore, and also that depending on artificial intelligence to resize photographs deprives a lot of control from users. According to the Twitter, not everything on the platform is supposed to fit greatly in the algorithm and in this particular case, only people can make a better judgment on how to crop images.
Twitter has increased the size of image views in its iOS and Android app to allow users to do the selection so the cropping decisions does not have to be made by algorithm. Whenever it is necessary to resize an image, such as when it is extremely tall or wide, the user will be provided a preview already of how it will going to appear.
The cropping algorithm is still being used, particularly on Twitter for the web, where image preview crops keep popping up, although Twitter appears to believe that in most circumstances, human judgment will take precedence over machine learning.
Source: Twitter.
Read next: A Phishing Scam On Twitter Takes The Appearance Of A DMCA Copyright Strike
Beginning in 2018, Twitter began cropping photographs to accommodate into its picture preview boxes using the algorithm. The algorithm struggled to find out what a user will find aesthetically appealing so it may incorporate it in the preview. However, some people criticized it for discriminating white faces over black ones. To verify if that is actually the case, Twitter asked some of its researchers conduct a study assessing its crop selection decisions.
On testing the algorithm for gender profiling, researchers discovered that it preferred white women to black women 7 percent of the time, and white males over black males 2 percent of the time. Overall, it has favored white people over black people 4 percent of the time.
Twitter additionally looked for gender discrimination, seeing if the algorithm crop photographs of women wrongly i.e. cropping in an inappropriate way around chest or legs rather than face. It discovered that though the algorithm prioritized women 8% of the time but the cropping is not done inaccurately. In roughly 3 percent of photos where it did not crop the face, other items were focused such as jersey number of sport shirt.
Twitter acknowledges that the algorithms may be damaging in aspects it may not have explore, and also that depending on artificial intelligence to resize photographs deprives a lot of control from users. According to the Twitter, not everything on the platform is supposed to fit greatly in the algorithm and in this particular case, only people can make a better judgment on how to crop images.
Twitter has increased the size of image views in its iOS and Android app to allow users to do the selection so the cropping decisions does not have to be made by algorithm. Whenever it is necessary to resize an image, such as when it is extremely tall or wide, the user will be provided a preview already of how it will going to appear.
The cropping algorithm is still being used, particularly on Twitter for the web, where image preview crops keep popping up, although Twitter appears to believe that in most circumstances, human judgment will take precedence over machine learning.
Source: Twitter.
Read next: A Phishing Scam On Twitter Takes The Appearance Of A DMCA Copyright Strike
