Twitch introduces a mechanism to detect unsolicited users who continue to troll by changing accounts no matter how many times they are blocked by machine learning


by

Marco Verch Professional Photographer

Twitch, a live game platform, takes further steps to reduce harassment cases and uses machine learning to detect people who harass or vandalize accounts or block them, Suspicious User Detection. Is reported to have been introduced.

Twitch will use machine learning to detect people evading bans --The Verge
https://www.theverge.com/2021/11/30/22810180/twitch-machine-learning-suspicious-user-detection-bans

Twitch distributors can block users from posting discriminatory or harassing comments. However, no matter how much you block, there are some who set up another account and continue to troll. The machine learning model that underpins Suspicious User Detection evaluates user behavior, account characteristics, etc., and compares that information with account data banned from streamer channels to identify the same person as the already blocked vandalism. Verify the possibility of being.

Messages from users who are determined to be 'probably blocked' will not be sent to the chat, but will be visible to broadcasters and moderators with channel management privileges. Distributors and moderators can also choose to continue to monitor or block the user in question, and if they continue to monitor, a message indicating monitoring will be displayed next to the target user name.



Twitch says Suspicious User Detection is on by default, but distributors can tweak or turn off this tool as needed. In addition, distributors and moderators can manually select and monitor suspicious users. Allison Huffman, Product Director and Community Health at Twitch, said, 'This tool was heavily inspired by the community's opinion that'we need a better way to deal with block evaders'.' 'It's a thing' statement to The Verge, an IT-related news site.

In addition, Huffman told The Verge, 'Whether users who make comments that violate the rules of the channel are repeatedly annoying harassment or new viewers who do not know the rules of the channel. That's why we designed Suspicious User Detection, which allows creators and channel moderators to make more efficient and informed decisions within the channel. It will be. '


by Brickin Nick

Distributors can also require users to join the chat to authenticate using their phone number or email address, and when combined with Suspicious User Detection, the difference between annoying users and new viewers. Will be easier to understand. However, The Verge says, 'It's quite possible to find out how effective detection of suspicious users really is, and how those who avoid blocking and continue to troll channels can avoid Suspicious User Detection. It also suggested that Suspicious User Detection may not be able to control vandalism.

in Web Service, Posted by log1i_yk