I agree with what you write here. I saw your exchange with @timcliff and was, frankly, dissapointed with his responses. I came away thinking either that (a) those making the fundamental decisions on Steem rule changes don’t really understand what is needed, (b) they get it, but are too busy protecting their profits, or (c) you and I don’t really understand what will work or not work. Steem IS too complex. As you write above, if you want people to use it, it needs to be simple and transparent. Steem is based on transparency of the block chain, but fails when it comes to transparency of the effects of user actions. It is also heavily stacked in favor of those with large stakes instead of those with quality content and interactions. Perhaps it is time to team up with others that share your views, such as @kevinwong and @scipio who had all proposed changes that appear to address some of the shortfalls in Steem.
That's funny because I couldn't be more opposed to Kevin Wong's proposed changes. They go directly against all of what I've written about here. Scipio's is much more interesting but I wish he would write about it more clearly. I'd be curious to hear @markgritter's opinion on it, as he's much more the graph expert than I am.
To be fair to Tim, curation is a fiendishly complex topic. I don't think Tim is actively out to hurt anyone, that seems entirely against his style. But he seems not to have put in the time to understand what this proposal actually does. Maybe if more of us write about it he will be inspired to do so.
From what I understand, Scipio is proposing a PageRank-style system in which rewards are based on the product of "user authority" (a reputation score derived from follower relationships) and SP. Leaving aside the technical challenges, this is probably still a "rich get richer" scheme. We could probably run it as an experiment across like a week of data and see what it would result in; I'm not sure I see any evidence Scipio has done that.
Even Google does not use PageRank-as-originally-described but has put a lot of work into link farm detection and other anti-spam techniques that do not fall out automatically of UA as Scipio claims it does.
The fundamental problem in assigning a base weight (the value of 1-d is greater than zero) to everybody is that there is literally no limit on how large a Sybil attack may be. Perhaps the account creation fee outweighs this, but it isn't obvious to me. But if cost of creating an account < benefit of a single UA follow from that account, spammers will create fake accounts. Whatever pattern of "real" follows is rewarded, a spammer could just create a copy of that and get an equal reward.
I think you can show that in the UA scheme, the net UA per user is about 1 (assuming no leaves that nobody follows) and all you can do is change the distribution, but a closed network can achieve that all by itself, and 1 UA per user times as many users as the spammer can afford could be very large. The spammer doesn't need to concentrate all that in one user to be effective, if sum of UA is used to calculate rewards. But that's just playing around with an Excel model, not solving the math. (If we view it as a probability distribution, the number 1 makes sense.)
For example, is there any reason to believe half of the accounts on Steem couldn't be spammer-controlled? Then they would get half of the UA. I think this applies to smaller fractions as well.
I was wrong about the leaves, they don't decrease total value, but popular accounts that don't follow anybody do create problems. Unfollowing can decrease your own value! Here's my spreadsheet model (1-10 are legit users, 11-20 is a spam farm)
User 4 is following 1, 2, 3 (they're part of a whale clique) and is sitting at 2.079 UA.
If user 4 unfollows just one or two of his clique, total UA is preserved and 4 stays about the same, but the remaining members of the clique go up.
But if user 4 unfollows all three, then value is actually destroyed. He's sitting at 0.923 instead, and the total UA for users 1-10 is 4.77 instead of 10.0. (I assumed self-follows were not allowed.)
User 4 can use this to his advantage. Say user 20 is a sybil controlled by him.
When 4 follows: 1, 2, 3 in this model he has 2.366 UA (the bump is from the blank user.)
When 4 follows just 20 instead, he has 3.785 UA and his sybil account has nearly as much, 3.367 UA. Not a very robust system at the small scales we can easily explore. A popular user can "defect" and capture a lot of the UA.
Isn’t this why we need one human = 1 account only? Oracles can do that, by there is no plan to change Steem to support it. Unclear why other than to support bots or special projects, but those could be treated like corporations with requests to create them.
I don't know of any online service that has successfully managed to implement such a rule. Maybe match.com? Identity checking is hard and spammers have ready access to fake credentials such as phone numbers.
And there're good reasons to have multiple account, or at least sub accounts. Eg keeping your blog on topic if you cover a range of things. I have plans to use steem for all sorts of things beyond just blogging and for that reason it is logical to have multiple accounts.
This could also be solved by UI changes to one of the front ends, perhaps combined with backend changes to support multiple channels from a single account. Ultimately, I think we need one human = one account AND one corporation = 1 account. With a solid process for identifying humans and a solid process for registering corporations, that would be possible. The blockchain would need to be modified to handle creation and destruction of corporations as well as "legal" oversight of both.
Yes, this is all complex because people can often find workarounds or ways to exploit the system that lead to unintended consequences. As for Kevin’s proposals, I’m not entirely sure of the all the details and have not thought through all the implications, but his basic idea that the current rules are incentivizing behavior that is not adding value and that we should try to change them to reward behavior that incentivized quality content and quality curation is spot on. We cannot expect people to just do the right thing. What is needed, as I’ve written elsewhere, is probably a way to experiment with different rule sets in parallel through something like opt-in subcommunities. Ned replied that SMTs + Oracles might enable this. With this approach we could enable competition among rule sets and also gain lost of experience in what works and doesn’t work. My only concern is that it may lead to an infinite regress where you would then have to have a layer to manage the creation of those communities in order to prevent the powerful from creating a community that benefited themselves.
Steem and Steemit are both open source. There's already at least one full clone of the system running out there. SMTs could help but it's possible right now to split off and run one under your own ruleset if you want to, either as a testnet or in actual production. The lower the price gets here, the more practical that becomes.