Twitch will launch an improved reporting and appeals process in 2022
Twitch will use machine learning to catch ban-dodging trolls
The tool will prevent ‘likely’ culprits from taking part in chat.
Twitch is introducing a new machine learning feature to help streamers protect their channels from people attempting to avoid bans. Dubbed “Suspicious User Detection,” the tool will automatically flag individuals it suspects may be “likely” or “possible” ban dodgers.
In cases involving the former, Twitch will prevent any messages they send from showing up in chat. It will also identify those individuals for streamers and any mods helping them with their channel. At that point, they can decide if they want to ban that person. By default, possible repeat trolls can send messages in chat, but they too will be flagged by the system. Additionally, Twitch says creators have the option to prevent them from sending any messages in the first place.
“The tool is powered by a machine learning model that takes a number of signals into account — including, but not limited to, the user’s behavior and account characteristics — and compares that data against accounts previously banned from a Creator’s channel to assess the likelihood the account is evading a previous channel-level ban,” a Twitch spokesperson told Engadget when we asked about the signals the system uses to detect potential offenders.
While Twitch plans to turn on Suspicious User Detection for everyone, the tool won’t automatically ban users for streamers. That’s by design because it’s impossible to create a machine learning tool that is 100 percent accurate in every context. “You’re the expert when it comes to your community, and you should make the final call on who can participate,” the company said in a blog post. “The tool will learn from the actions you take and the accuracy of its predictions should improve over time as a result.”
The introduction of the tool follows a summer in which Twitch struggled to contain a phenomenon called “hate raids.” The attacks saw malicious individuals use thousands of bots to spam channels with hateful language. In many cases, they targeted creators from marginalized communities. Hate raids became such a frequent feature of the platform that some creators walked away from Twitch for a day in protest of the company’s lack of action.
(26)
