Twitter is announcing a global change to its ranking algorithm this week, its first step toward improving the “health” of online conversations since it launched a renewed effort to address rampant trolling, harassment and abuse in March. “It’s shaping up to be one of the highest-impact things that we’ve done,” the chief executive, Jack Dorsey said of the update, which will change how tweets appear in search results or conversations. “The spirit of the thing is that we want to take the burden off the person receiving abuse or mob-like behaviour.”
Social media platforms have long struggled to police acceptable content and behaviour on their sites, but external pressure on the companies increased significantly following the revelation that a Russian influence operation used the platforms in coordinated campaigns around the 2016 US election. Facebook and Google have largely responded by promising to hire thousands of moderators and improve their artificial intelligence tools to automate content removal. Twitter’s approach, which it outlined to reporters in a briefing on Monday, is distinct because it is content neutral and will not require more human moderators. “A lot of our past action has been content-based, and we are shifting more and more to conduct,” Dorsey said.
Del Harvey, Twitter’s vice-president of trust and safety, said that the new changes were based on research that found that most of the abuse reports on Twitter originate in search results or the conversations that take place in the responses to a single tweet. The company also found that less than 1% of Twitter accounts made up the majority of abuse reports and that many of the reported tweets did not violate the company’s rules, despite “detracting from the overall experience” for most users. The new system will use behavioural signals to assess whether a Twitter account is adding to – or detracting from – the tenor of conversations. For example, if an account tweets at multiple other users with the same message, and all of those accounts either block or mute the sender, Twitter will recognise that the account’s behaviour is bothersome. But if an account tweets at multiple other accounts with the same message and some of them reply or hit the “heart” button, Twitter will assess the interactions as welcome. Other signals will include whether an account has confirmed an email address or whether an account appears to be acting in a coordinated attack. With these new signals, Harvey explained, “it didn’t matter what was said; it mattered how people reacted.”
The updated algorithm will result in certain tweets being pushed further down in a list of search results or replies, but will not delete them from the platform. Early experiments have resulted in a 4% decline in abuse reports from search and an 8% drop in abuse reports in conversations, said David Gasca, Twitter’s director of product management for health. This is not the first time that Twitter has promised to crack down on abuse and to troll on its platform. In 2015, then CEO Dick Costolo acknowledged that the company “sucks at dealing with abuse and trolls.” But complaints have continued under Dorsey’s leadership, and in March, the company decided to seek outside help, issuing a request for proposals for academics and NGOs to help it come up with ways to measure and promote healthy conversations. Dorsey and Harvey appeared optimistic that this new approach will have a significant impact on users’ experience. “We are trying to strike a balance,” Harvey said. “What would Twitter be without controversy?”
Credit: Julia Carrie Wong for The Guardian, 15 May 2018.