One problem with tagging is how much false tagging bullshit happens. You can't trust the posters to put correct tags, but you cannot also expect malicious users not to put false tags to hide stories from others.

I've also always hated binary up/down voting systems. Slashdot had it better with meta moderation where you had a few options to choose from.

I suppose now with AI I could mock up a UI concept I call orange slice voting. Instead of a singular up/down vote, you get what looks like a orange sliced across its equator where each segment has a series of positive and negative vote options and the user gets one selection per post.

"I like this content", "I believe this is true", "Fits this thread", "Good post", and "Misinformation", "I don't like this content", "doesn't fit this thread", "etc"

These can be adjusted for a site as needed and gives more dimensions for people to search and filter by.