RE: Specialization - jump starting new steemit sub-communities and creating insurmountable obstacles for others
Once a new category is established, it would be nice if there was a category fund that can act as a virtual whale for that category. I can think of automated ways where if enough high reputation accounts are upvoting a post in a specific category, the virtual whale upvotes. This automation would put the power in the hands of the dolphins and relieve the whales from having to work 24/7. Also by the whales being less active, their stake is gradually diluted, which helps distribution.
You are apparently not thinking about the game theory of how this can encourage Sybil attacking it for gains. There is a reason the whales are given control with a quadratic weighting, and it is because otherwise all sorts of ways to game the voting system are enabled.
It isn't going to be so easy to fix Steem's voting and ranking algorithms, because it is a fundamental problem.
the elected curators upvotes are just tallied, so not sure how that can be sybil attacked. a sybil(s) would need to be elected as a specialist curator. It is more a way to separate the voting power from the voting decision.
Given that a majority of specialist curators need to upvote for the virtual whale upvote, what attack vector is there?
The assumption is that any sybil curator that somehow gets elected would be detected and unelected. Clearly if the specialist curators are all infested with sybils it wont work, that is why I said that the curators need to be known accounts with knowledge of the speciality. How can a sybil convince the community they are knowledgeable about a topic and continuously pass scrutiny by the other specialist curators that they are not some sybil?
If by “elected curators” you mean what you wrote previously “if enough high reputation accounts are upvoting a post in a specific category, the virtual whale upvotes”, then it means one only needs attain the threshold of upvotes (reputation) with their accounts and use this to influence rewards. Remember voting reward is non-linear, so if you can mount a swarm attack that mobilizes the whales' voting power (per your suggestion to apply their voting power algorithmically), then you can cheat the system. And if you make the weighting algorithm non-linear, then the entire system is ruined because everyone has an incentive to vote for themselves.
if you read the discussion the process is clarified. let us use #trading as a specific example. We would find half a dozen people (regardless of how much SP they have) that are knowledgable about trading to be the specialist curators. The exact method of how they are selected is not finalized and certainly subject to a continuous review as it is quite important.
So let us assume there are N such specialist curators. Now a majority of these curators need to upvote a post with a #trading tag for it to trigger the autowhale upvote.
People are involved in the entire process, the only thing automated is tabulating the specialist curator's votes and trigger the autowhale vote. Presumably if a sybil account can fool the community that it is knowledgeable about #trading and it is making votes consistent with such a person, then I claim there is actually no difference between that sybil and a real person, hence there isnt a valid sybil attack as the result is the same.
Please comment on what is proposed, not what you assume is proposed.
I do not see any sybil attack, short of an AI that can pass an ongoing Turing test
I commented on what was proposed in the blog post. I didn't read your comments after you made the blog post.
This seems to be different than what you proposed in the blog post. When you wrote “reputation” in the blog post, it presumably means the reputation system recently implemented which is that number in parenthesis next to our username which is tabulated from vote history not elections.
Okay so now you morphed (or clarified) your proposal to elections of delegates who will control (some portion of) the whales' voting in the instances the majority of them (a quorum) agree.
There are some issues with this:
Radically improving relevance will be a major breakthrough. I don't think your proposal will be that significant of an improvement because it lacks algorithmic power to develop emergent phenomena in relevance and like-mindedness, although it might spread rewards around a little bit better (unless #1 is entirely gamed as it is always is in politics due to the Iron of Political Economics and the power-law distribution of wealth).
specialist curators will need to respond to questions about their votes (within reason) so we can make sure it is not a bot, ie. it can read, understand and respond like a human.
If the sybils that are run by an attacker somehow gets the majority of the specialist curator spots, then it would be up to the community to escalate things to a higher level review of voting pattern. Since all the votes are public, we can see what posts that should have gotten votes didnt and which ones that shouldnt have did. Given that information, and a removal process of a bad curator, I think any damage will be time limited
replying to your other reply.
The aim for my plan is not to make spontaneous identification of new cateogories, but rather to allow the community to select specific high value specialist categories, fund it, and create a reasonable enhancement to posts in that category.
I have not specifically stated as I thought it was clear, but all posts in the specialist categories still go into the main steemit system and can get the upvotes via the normal method. The problem I am trying to solve is where new users arrive all excited, spend a lot of time creating valuable content and get $1.39, feel disappointed and leave. This robs the community of a source of good content and increases the dropout rate, both are problems that need to be fixed without needing to conflate other problems that also need to be fixed onto this one.
By having dozens or hundreds of specialist categories where a good post has its revenues enhanced via whatever flawed political process, is still better than $1.39 for a high value content