Twitter has launched a reward challenge in which the system pays a reward to anyone who finds a bias when it automatically crops an image. Compensation is usually a bug bounty program, in which money is donated to someone who finds a security problem or a flaw in the process. However, this time, Twitter is running a contest to find out if there is any bias in the algorithm that crops and crops to a good-looking state in order to make thumbnails of user posted photos.
Twitter’s automatic image cropping started in 2018, but some users have criticized the algorithm for tending to perform biases such as centered on white-skinned people. In May, Twitter paused its image cropping algorithm offering, shared how to identify perception bias, and released code to reproduce it. It also said that cooperation is needed to go further and identify potential problems.
According to Twitter, the contest is the industry’s first reward program targeting algorithmic bias. Prizes are up to $3,500. Twitter says it wants to change machine learning model biases because it is difficult to find and unintended ethical issues can be found. He also explains that the reason for implementing this program is because we believe that we need to identify and compensate for these problems and because we cannot solve these problems on our own. The contest is open for registration from July 30 to August 6, 2021, and the winner will be announced at the event (DEFCON AI Village) on August 8. Related information can be found here.