Facebook ‘relying on users to report most abuse on the site’

FACEBOOK’S technology is detecting just a tiny proportion of the abusive posts by bullies and trolls, it has emerged.

The firm revealed it had taken action against 2.1million abusive posts in three months – but just 15% were detected by Facebook, with the company relying on users to report the rest.

Artificial intelligence and algorithms fail to understand nuances in posts which might mean a message is abusive, Facebook admitted.

A ‘transparency report’ published this week contained data relating to bullying for the first time.

Online trolling could be ‘especially harmful to minors’, Facebook admitted. But the firm said: ‘Bullying and harassment are highly personal by nature. In many instances, we need a person to report this behaviour to us before we can identify or remove it.’

The report comes only days after Facebook boss Mark Zuckerberg snubbed a request to appear before an international ‘grand committee’ to answer questions about ‘disinformation and fake news’. Ireland was one of five countries supporting the invitation, led by British MP Damian Collins, who demanded Mr Zuckerberg appear in Westminster on November 27.

Fine Gael’s own members will put pressure on the Taoiseach at this weekend’s ard fheis to bring in a Digital Safety Commissioner with ‘real teeth’ to police social media sites such as Facebook. The debate is likely to cause embarrassment for Leo Varadkar, who indicated in July he was in favour of selfregulation, as he said a ‘Digital Safety Commissioner in Ireland will not be able to regulate the world wide web’.

Mr Zuckerberg said during a company call on Thursday night: ‘We are going out of our way to proactively enforce the policies using this combination of AI systems and just dramatically growing the review team.’

But Facebook may never be clean of horrific content, he conceded.