AI moderation is a confidence trick

AI moderation is a confidence trick

Hot take: platforms talk up their 'advanced AI detection' precisely because it mostly doesn't work, and they need you to believe it does. The fear of the system is doing the enforcement the system can't.

If the model were as good as the press release, they wouldn't need the press release. Deterrence is cheaper than detection. A scary blog post about machine learning catches more 'violators' than the actual classifier, because half of them self-censor out of dread.

You're being moderated by a rumor.

Agree? Fight me in the comments.