Twitter is laying out details of its bias bounty as it challenges the public to sift through its code and find ethical errors in its image-cropping algorithm. The contest comes as it was discovered a while back that the algorithm might have a slight bias against black people, something that the company is trying to eradicate by testing uncropped image previews instead.
Twitter started using “smart cropping” in 2018, but users claim that it tends to favour light-skinned people and even women over men. The saliency algorithm uses machine learning to crop the images so that previews in tweets remain a consistent size. Now, the company is seeking help to find the bias in its machine learning model before it causes more “unintended ethical harms.”
For the coders who are interested, there are three cash prizes ranging from USD$500 (~RM 2,112) to USD$ 3,500 (~RM 14,789) and two special cash prizes worth USD$ 1,000 (~RM 4,225) for innovation and generalisability (being applicable to more algorithms). Winners will be announced at the DEF CON AI Village workshop on 8 August and the winners will present their findings to the public.
“In May, we shared our approach to identifying bias in our saliency algorithm (also known as our image cropping algorithm), and we made our code available for others to reproduce our work,” Twitter wrote in a blog post. “We want to take this work a step further by inviting and incentivising the community to help identify potential harms of this algorithm beyond what we identified ourselves.”
Rumman Chowdhury, director of Twitter’s Machine Learning Ethics, Transparency and Accountability team, tweeted that the reason for the bounty is because “we believe people should be rewarded for identifying these issues, and we can’t solve these challenges alone.” The contest opened for entries on 30 July and will only run until 6 August.