Sorry for the delay,
I don't really thing anonymous downvotes are this bad, proof of brain in theory doesn't require you to know who is voting/downvoting you, it's like likes/dislikes on social media.
I believe what you suggested with tags has already been though of and even tried but I don't remember when. I think it was abandoned because people people would bot the service to get downvotes to people they didn't like even if the post was good.
I think curangel has something like this where people vote on posts to be downvoted.
But it's an interesting one, having a layer 2 service where people can vote on whether a post should be downvoted or not anonymously and then a bot on which people would delegate voting power to it would downvote depending on the results of such votes.
But downvoting is always tricky and causes a lot of frustration from users and downvoters, I honestly think that while proof of brain is great on paper, it fails in real life and we should try to find another model for reward distribution altogether.
Do you have any thoughts or ideas for what that model might look like?
Not yet but it's thoughts I've been having, I need to take some time to think this through.
Cool, keen to hear when you've had some chance to think more about it.
One thing I myself am quite interested in is allowing communities to experiment with different economic models. I'm not talking about SMTs, although of course they will come later and allow even more experimentation. But what if a community owner account can be set as the mandatory beneficiary of 100% of the rewards generated in that community? Maybe those who are solely interested in the rewards will not participate, or maybe they will, depending on what the community decides to do with the rewards (e.g. a local group can have monthly parties paid for by the community's rewards, etc.). How will people's behavior in that community be different if they don't personally seek to maximize their rewards or even participate for the rewards? Hive without the rewards seems like an interesting experiment which I think can easily be tried with some additions to Hivemind.