This started off as an article. It then grew into something like a mini whitepaper. I then thought that most people would prefer to read a brief summary. So, in the end, I just wrote the summary! This is what you will find below. Remember, these are just ideas for discussion, with the best interest of the whole community.
Summary of Ideas for Future Rule Changes
The Steem blockchain is a game theoretic construct. Its rules are encoded in the rewards algorithm. There is no blockchain without rules.
The primary objective of the rules should be to maximise social benefits by maximising the sum of individual benefits.
The rules must evolve and become more complex so that each iteration of the rewards algorithm should converge towards the aim of maximal social benefits.
Humans are complex and therefore will adapt their behaviour to maximise their own personal gain - this cannot be totally eradicated from the game.
The so-called Prisoner's Dilemma is a very real phenomenon in social behaviour. Put in terms of social interactions, it states that the equilibrium point is the point at which everyone acts in their own self-interest. But such an equilibrium point is not the optimal point. However, for any one individual to change behaviour unilaterally from one of self-interest to one of optimal social interest is not in the individual's own self-interest. That is the crux of the dilemma.
Just to drive this point home: the sum of individual benefits is lower in the Nash equilibrium than it is in the optimal scenario.
The best that can be achieved, in the game theoretic sense, is to make optimal social behaviour more attractive than the sub-optimal equilibrium point. This must be encoded in the rule-set, the algorithm, and cannot be expected from mere behavioural changes because the Prisoner's Dilemma is a powerful disincentive.
Blockchains are designed to be trustless interactions meaning that the Nash equilibrium point will prevail over the optimal point unless encoded in the rule-set.
The components of a rule-set should not be looked at in isolation but rather as a whole. It is the social interactions, the feedback loops, that govern the success of the whole rule-set towards achieving the optimal social outcome. The balance between rewarding rules and punitive rules should always be to increase the likelihood of such an optimal environment.
Each iteration of the rule-set will create short-term instability until a new normality prevails. This is just the nature of complex systems. Hopefully, each iteration will increase the probability of achieving the primary objective.
Now we can look at some concrete proposals. As Hard Fork 20 (HF20) has already been planned, these could be considered for HF21 or beyond.
The reward curve should remain proportional to R, the number of rshares associated with each post or comment.
Self-voting and voting cliques cannot be eliminated. Indeed, for curators self-voting can be important in triggering votes from their followers. It is also a waste of social power to spend considerable time on negative interactions at the cost of more positive ones. Therefore, there can be a rule that voting for the same user, whether oneself or another, will decrease the power of each subsequent vote, within a limited period. For example, take the last 10 votes of a user and, if a new vote is given to a user already voted then that new vote may be worth 90% or 80% of what it would normally be. This can be scaled down to the point that further votes have even less value. I repeat, this will not stop such behaviour but will decrease the reward pool that it generates and therefore mitigates some of the effects on the whole system.
The good thing about algorithms is that they are predictable. The bad thing about algorithms is that they are predictable. Life has a lot of randomness in it. We cannot wholly predict the future and there are both good and bad surprises that happen every day. Surprises in terms of the rewards algorithm means that a small proportion of votes will see either an increase or decrease in value. This is one simple technique to, again, incrementally shift behaviour towards an optimal outcome. This will increase social interaction and encourage meaningful communication with multiple other users. This may be counter-intuitive but combined with (13) has the effect of increasing earnings through interaction rather than treating voting as a kind of mining technique. That can still be done with adjusted income expectations.
The drain on voting power should be decreased from 2% to 1%. It should also be clear in the User Interface (UI) such as Steemit what one's remaining effective voting power is so that newbies do not drain their power by accident. Voting at 100% power, the 2% drop per votes, and the 20% replenishing within 24 hours, means that each user has just 11 votes per 24 hours without losing power. Voting at 100% power and a 1% drop per vote results in 23 votes per day. If a user decreases their voting power to 50%, then the 2% and 1% drops yield 23 and 45 votes respectively. But minnows only have the voting slider within chainBB and not on Steemit. As mentioned by @aggroed (in an article by @timcliff), minnows especially are engaging less and somehow hording their power hoping for worthy content. It also increases the percentage of self-voting as there are then few votes left to distribute to others. I do not think the change to a 2% drop in power per vote has improved the primary objective and I hope this can be changed as soon as possible. There will be a drop in average earnings per vote but that just means each individual is interacting with more people.
The formula to calculate rshares must be then adjusted so that (15) does not drain the reward pool - I'm not sure why this part of the rule-set was not adjusted at HF19.
I think that is long enough for now! Your comments appreciated and please read the entirety of the article as some sections interact with other sections. I have numbered them for ease of reference and not their relative importance. Any glaring errors, please point them out and I will amend the text.
Thanks for reading!
Please leave a comment.....