UKTN features Arwen's work using AI to combat online hate speech

"Social media has become an invaluable tool for brands, but it comes with the inescapable reality of toxic engagement.

Spam comments, hate speech and other harmful content can create negativity around a brand’s social media posts and contribute to a worse digital experience for everyone.

While there remains an ongoing debate around how best to deal with potentially harmful digital content, fuelled in part by the development of the Online Safety Bill, one UK startup is offering a potential solution.

Surrey-headquartered AI startup Arwen has developed an algorithmic solution to toxic engagement on social media posts.

“Advertisers need brand safety. They need a safe environment in which to put their content, otherwise, they’re just going to get inundated with really horrible commentary,” Joel Bailey, product and service developer at Arwen told UKTN.

Platforms like TikTok, YouTube and Facebook have all implemented community guidelines that, to some degree, attempt to automatically find and remove harmful posts and comments. However, the volume of user content makes it an almost impossible task to get completely right.

“The networks are obviously global, they operate across multiple cultures, the volume is enormous,” Bailey said.

“I’m very sympathetic to the networks and the struggle, they have to make this work. And when you think about 600,000 comments being made per minute on Facebook, just in the UK, very quickly the scale and the data processing required becomes enormous.”

Arwen is taking a different approach at the individual and brand level in the form of bespoke content filtering for brands and public figures.

The company holds a consultation with customers to determine what types of comments and types of language come under what Bailey described as the “lawful but awful” category.

Arwen has 25 different algorithms, each one designed to detect different types of potentially damaging content, which include spam, profanity and insults, hate speech and identity attacks and more.

Arwen’s software automatically filters out this content, with brands and public figures paying a subscription fee for continued access to the service. It has been authorised by Twitter, Facebook, Instagram, YouTube and TikTok to plug into the respective social media platforms and use AI to remove unwanted content.

The company was founded in 2020 by Matt McGrory, Dr David Cole and Joel Bailey.

One of Arwen’s earliest customers is comedian Rosie Jones, who has faced ableist and homophobic abuse online throughout her career.

“Rosie was a very early customer of ours,” Bailey said. “Her point was protecting me is important, but it’s more important to protect my community because there’s lots of junior [versions of me] in my community and if they see people getting away with calling me something unpleasant, then that really demoralises them”.

Jones recently made a documentary on the abuse disabled people face online, which aired last Thursday.

The documentary showed Jones’ work with Arwen in filtering hateful speech in digital spaces.

“She [Rosie] called me and said we were thinking of doing a TV programme, it’s all part of the mission of making social media, more social, so we were up for that. She’s also our favourite customer, of course,” said Bailey."

Read the full article here https://www.uktech.news/ai/arwen-ai-hate-speech-20230724