Roblox Is Using AI To Moderate Online Chats
Credit: Roblox Corporation
Roblox has begun using an AI-powered algorithm to correct the content of user chat messages. This real-time rephrasing of messages goes beyond the platform’s existing filtering of banned words, which are replaced with “#” symbols when detected.
The main new feature of the mechanism was described as automatic rephrasing. Previously, expletives and phrases in the user’s original message were censored with the symbols “####,” but now the new filter automatically rewrites the entire text. Only the edited version reaches the recipient, and all game chat participants receive a notification about a detected attempt to violate the platform’s rules of polite communication. The company also announced that it is developing methods for recognizing disguised obscene expressions. For example, those in which some letters are replaced with numbers or punctuation marks; the overall meaning is still discernible. According to the administration, the number of false positives for expletives when posting or requesting messages has now decreased by 20 times.
The automatic rephrase feature for online messages on the Roblox platform is currently available in in-game chats between players who have passed age verification and are in similar age groups.
