How do you think AI comment moderation should work?

Question

Reading comments one by one and checking if they are in a way breaking the rules, is a boring and time taking (not individually, but because there will be lots of comments) task. We can use AI to do it faster. But, AI always makes mistakes. So do you think is it better that:We approve whatever AI approves without manual check and manually check if the comments rejected by AI are actually breaking the rulesOrWe manually check whatever AI approves, and use AI only to reject the bad comments?

Hugsbox · Accepted Answer

Don't give AI the ability to unilaterally remove/reject comments. Allow AI to flag comments as potentially harmful, and then have a human review whether or not they should be removed. AI shouldn't have the agency to actually do anything, just the ability to say "Hey, this comment might be against the rules, you should have a look!"