Twitter Removes Image-Cropping Algorithm After Finding BIAS in it

Image-Cropping Algorithm

Twitter said on Wednesday that it would discontinue its automatic image-cropping system after an investigation revealed bias in the algorithm that controlled the function.

The messaging platform stated it discovered “unequal treatment based on demographic disparities,” with whites and males receiving preferential treatment over Blacks and females, as well as “objectification” prejudice that focused on a woman’s breast or legs, dubbed “male gaze.”
The announcement comes only a month after Twitter announced the commencement of an algorithmic fairness program to decrease bias on its platform caused by automation.

In 2018, Twitter implemented a cropping algorithm known as the saliency algorithm, to improve uniformity on the size of photographs in one’s timeline and focus on their most significant features.

According to Rumman Chowdhury, the head of Twitter’s software engineering and an expert in ethics and artificial intelligence, the firm decided it is best to leave cropping choices to users after its research.

In a blog post, she stated, “We examined the tradeoffs between the speed and uniformity of automated cropping and the possible hazards we found in our research.”

In 2018, Twitter implemented a cropping algorithm known as the saliency algorithm, with the goal of improving uniformity on the size of photographs in one’s timeline and focusing on their most significant features.

According to Rumman Chowdhury, the head of Twitter’s software engineering and an expert in ethics and artificial intelligence, the firm decided that it is best to leave cropping choices to users after its research.

In a blog post, she stated, “We examined the tradeoffs between the speed and uniformity of automated cropping and the possible hazards we found in our research.”

Keep Visiting DigitalGyan for more Technical News and Updates.