So this is a good illustration of the problem.

If it were my site, "I like X now" would be a red flag.

I don't think you're gonna AI your way out of this part of things for some time, and it really is the core challenge to content moderation; it's heavily opinion and circumstance based, in a way current models really struggle with.

I appreciate the comments, thanks.

Well, we are going to give it a try!

Thanks again...