Twitter offers ‘good’ money to find algorithmic bias

TWITTER said last Friday it would provide "good" money to users and researchers to help eliminate algorithm bias on social media. ...


TWITTER said last Friday it would provide "good" money to users and researchers to help eliminate algorithm bias on social media.

The San Francisco technology company said it would be the "first algorithmic bias bounty competition", with prizes of up to US $ 3,500.

The competition is based on "bug bounty" programs offered by other websites and forums to find safety holes and be at risk, according to Twitter executives Rumman Chowdhury and Jutta Williams.

Finding bias in machine learning models is difficult, and sometimes, companies find out about unintentional behavior once they have already reached out to the public, "they wrote in a blog post.

"We want to change that."

They said the hacker bounty model offers the promise of gaining algorithmic bias.

"We are inspired by the way in which the research and non-research communities have helped the security sector to develop better ways to identify and reduce the risks to public safety," they wrote.

"We want to build a just society.

The move comes amid growing concerns about automated algorithmic systems, which, in addition to neutral attempts, may include discrimination or other forms of bias.

Twitter, earlier this year introduced the algorithmic editing program, said in May it said it was completing an automated image removal program after the discovery of its update to the task-algorithm.

The messaging platform claims to have found that the algorithm introduced "unequal treatment based on discrimination", with white and attractive men than black men and women, as well as "contradictions" based on a woman's chest or legs, defined as "male perception". AFP

No comments