Blizzard is testing the use of machine learning to address inappropriate language in Overwatch. This is done in different languages. In the long term, the technology must also be able to assess more than just language, such as behaviour.
Jeff Kaplan, game director of Blizzard’s Overwatch shooter, says in an interview with Kotaku that the company is experimenting with machine learning and trying to teach a system what unwanted language is. The purpose of betting ai is to be able to tackle this more quickly, without having to wait for a player to report it. It involves language use in different languages, such as English and Korean. Currently, Blizzard is deploying the system to deal with the most egregious cases.
“With anything to do with reporting and penalizing players, you have to start with the most extreme examples and see how the rules can be adjusted from there,” Kaplan told the site. The detection of inappropriate language would not analyze direct messages between friends. In the long term, detection of undesirable behavior in the game should also become possible. It is unclear how far Blizzard is with that. Kaplan says, “That’s the next step. For example, how do you know if Mei’s ice wall was put down in the spawn room by a jerk?”
Furthermore, the Overwatch team is looking at opportunities to reward positive behavior in the game. For example, together with companies such as Twitch and League of Legends maker Riot Games, it is part of the so-called Fair Play Alliance, which works on ‘healthy communities’ in online games. For example, in LoL, such a system already exists in the form of the Honor system. The use of machine learning for moderation is also happening in other places, such as the comments under New York Times articles based on a Google service called Jigsaw.