Automated moderation tools are flagging “Luigi” as potentially “violent.“

  • Propheticus@lemmy.zip
    link
    fedilink
    English
    arrow-up
    1
    ·
    2 days ago

    That’s what happens when you train your model by only feeding it positive cases. If you just count the times a word was involved in a post where a human mod banned/deleted, but not the times it was not… Then again how often were people discussing the game character outside of game related subs? Could simply be some truth to it when loads of people use the ‘to Luigi someone’ form.