Twitter will begin using a wider range of signals to classify tweets in conversations and searches, hiding more responses that are likely to be abusive the company said today . Comments from users who have often been blocked, silenced or reported for abuse will be less visible throughout the service, CEO Jack Dorsey told a group of reporters. "We are moving forward as we move forward," said Dorsey.
Twitter already classifies the tweets in the search and in the conversations. But until now, it has not taken into account the negative signals when classifying them. This has meant that the answers could be easily evaluated by the bad actors, whether they are spammers selling cryptocurrencies or networks of robots that try to influence the elections.
Twitter will begin to examine a much wider variety of signals when ranking tweets in conversations and in search, Dorsey said. Some of these signals include the number of accounts created by the person who tweets, the IP address and whether the tweet has led people to block the person who tweets it. Twitter will not remove Twitter tweets, he said, but will now move to the "see more answers" section of a conversation, where they are hidden behind an additional touch.
A test of the new approach to the ranking found that the number of reports of abuse generated by the talks decreased by 8 percent, the company said. "The spirit of the matter is that we want to take the burden of work away from people who are being abused or harassed," Dorsey said.
Relying on algorithmic signals could have several advantages for Twitter, since it works to reduce abuse on the platform. They work without respect to the content of the tweet, preventing Twitter from having to make difficult decisions about the tone or intention of a message. And they work independently of the language in which the tweet was written, which allows the company to distribute the changes globally at once.
At the same time, decisions made by algorithms can also go disastrously wrong, and can be difficult for outsiders to understand. Dorsey said Twitter is aware of that and would invest in making sure the product communicated about how it makes decisions. The company will also consider issuing reports on compliance measures it adopts through the platform, said Del Harvey, vice president of trust and security for the company.
The moves are meant to address Twitter's long-standing struggle to control abusive accounts. The company has made several high profile moves in the last two years to make the service feel more civilized, but it continues to generate criticism for its opaque policies and inconsistent application.
The changes, which will be extended to all users worldwide this week, come two months after Twitter issued a request for proposals from researchers and academics to help the company measure the state of public conversations in the place. The company received 230 proposals of this type, said the product manager, David Gasca, and Twitter will announce the next steps with these proposals in the coming weeks.