In an attempt to reduce the toxicity of Twitter, the social network has created a new alert system that detects when a message brings elements considered offensive. After a test period, this feature started working on Wednesday (5) for all users who use the English version of the service.
When Twitter’s system detects that the language in a tweet may be offensive or hurtful, it prompts the user to review what they have written, giving them a chance to change their mind. In its official blog, the company explains that its algorithms take into account the relationship between the person who created a message and the person who is replying to it so that jokes are not included in its filter.
Twitter claims that testing from late 2020 helped it detect nuances more accurately and cut down on times when warnings appeared inappropriately. “The tests resulted in fewer people sending potentially offensive messages through the service and improved behavior on Twitter,” the company said.
The system has already brought good results
According to the company, the new system has caused 34% of affected users to either stop sending insults or change their messages, resulting in an overall 11% drop in the number of messages deemed inappropriate. In addition, those exposed to an alert were less likely to repeat their behavior and less likely to deal with further warnings suggesting a behavior change.
Looking ahead, Twitter says it will continue to study new methods of intervention to ensure a healthier environment for its users. In addition to alerts for offensive messages, the social network also encourages users to read the content before sharing it in an attempt to curb the spread of potentially harmful fake news.
Source: Twitter Blog