In the past few months, Twitch streamers have been dealing with an increase in hate raids, especially for those in marginalized communities. However, a newly-revealed tool hopes to use machine learning to help protect streamers and catch users trying to evade channel-level bans.
Announced yesterday in a blog post, Twitch talks about how a common tactic among bad actors is to create new accounts to continue their harassment after getting banned. The new tool, Suspicious User Detection, uses machine learning and account signals to help users catch these users based on "account signals." When users have the tool turned on, chat messages may be flagged as "likely" or "possible" ban evaders. "Likely" messages are visible only to the streamer and moderators, while "possible" messages are visible to all but flagged for the streamer and moderators. From there, the streamer and their moderators can decide on whether the message is okay or if it needs deleting and if the user should be banned.
Suspicious User Detection, powered by machine learning, is here to help you identify and restrict suspected channel ban evaders from chatting before they can disrupt your stream.— Twitch (@Twitch) November 30, 2021
Learn more here: https://t.co/01cCwnQZfw pic.twitter.com/QWVSnRPg1X
Twitch notes that, particularly around launch, that no machine learning will be 100% accurate. Streamers are generally the experts on their own community and have to make the final call on what messages go through. However, the Suspicious User Detection tool will give streamers better control over their community and will learn over time based on user input, making it more accurate over time. So far, public response to the new tool has been positive, with RekItRaven, one of the main actors behind a Day Off Twitch, thanking Twitch for continuing to introduce new features to keep streamers safe.
This isn't the only tool Twitch has rolled out in order to help stem the tide of hate raids. Back in September, they gave users the ability to require chatting users to verify their accounts with their phone number and/or email.