Curation needs to be more competitive with self-voting if we wish to game-theoretically incentivize socially-beneficial behavior. That may be difficult, but perhaps we can at least economically disincentivize the "bandwagon-curator" effect.
One of the core problems cited again and again by Steemit users is that many posts make high rewards based mostly on the assumption of high curation value, rather than on quality. Often, the "curators" haven't even read them, or aren't even physically capable of doing so, as they are bots (automated). Their point is well taken and nigh impossible to contest. This currently seems to be the optimal way to maximize one's curation rewards per unit of voting power, short of direct self-voting.
What we need to devise is a way to counteract the bandwagon effect that the current curation formula creates without punishing those users who legitimately post very high value posts. In other words, the maximal curation incentive should be supplied to the objectively highest quality content (of whatever niche), not by up-voting a digital Jersey Shore.
Further, if we are careful to stick to the scope of the curation issue, the problem is not how much the author is receiving in rewards, but the fact that curation is currently rewarded primarily by how much the author receives in rewards.
Ending the situation where rewards total is the sole factor in curation reward would allow attacking the curation problem without attacking the author, and hopefully quality would rise to the top.
In fact, an excellent system would make it highly profitable to unearth hidden gems on Steemit via curation.
I have a few suggestions for ways to start, and hopefully the community's programmers can take them and do much better. Note that when I use the term "target", I refer to the author of the post or comment being upvoted to curate.
This is the simplest to implement. You simply cap the amount a user can earn from curating a target, either by a flat amount or a formula relative to any number of factors. You could restrict curators to a "1% or .1% of total target author's lifetime rewards", or simply "X Steem Power per account". Any amount decided upon could be modified by curator Steem Power.
The logic here is that once an author has generated a certain amount of curation rewards for a user, they have probably already received the bulk of any other views/followers the curator will generate. Further resteems by this curator will go to the same subset of that curator's followers, leading to a decreased likelihood of exposing the post to new users since they may already be following the target from past resteems (or organically.)
Since the bandwagon effect will be reduced, the amount of curation rewards rewarded to legitimate curators should be increased. If necessary, a greater share of rewards for curation could be considered once the system is less game-able. Even going from 75/25 to 70/30 is quite material.
Being an arbitrary rule, this is going to occasionally work very poorly, unless a favorable formula can be found for a moving cap. Is the cure worse than the disease? However, one advantage this approach has is the requirement to be powered up to curate; this option can only be minimally gamed by sock-puppeting or delegation.
Alternatively, curation rewards could be capped by target account instead of by curator. In other words, once a certain number of curation rewards are paid out for a particular author's work (either total or time-gated), no more can be received.
Implement a second reputation score for curation. This could remain hidden to the user by default, like NSFW posts, or until a certain threshold score is reached, to avoid further bloating the interface for new users. Modify curation reward based on this score to encourage long-term thinking for curating quality.
The logic here is to allow far higher rewards from curation than would be feasible (due to abuse), but gate it with a higher curation reputation level. If implemented, curation reputation could be started at 25 for each account, or derived from other account factors such as reputation and age.
This would only work with some tweak to the curation formula currently in use to avoid simply furthering the bandwagon affect. If combined with option 1, a curation reward cap per target author, it could encourage far ranging curation to maximize curation reward. That reward could be increased at high curation reputation levels to justify the effort relative to self-voting game theory, which remains the baseline in terms of lowest effort.
Again, with this additional control on the curation system, material changes small enough for authors to support could be made to the author/curator split. 70/30 may be enough, 66/34 is the next step.
Slapshod alternative: Modify curation rewards by standard reputation (higher/lower reputation are both possible.) Reputation modification (adjustment/reset) will be controversial if deemed necessary.
In part two, I'll discuss my next several suggestions: modifying curation rewards inversely to author statistics, stake-weighted silent black-lists, and voting power refunds.
Thanks for reading, and please let us know your thoughts in the comments. In particular, if you have an idea for rewarding curation more without detracting from author rewards, I'm sure we'd all love to hear it!
Sources (part 1 + 2): KeepCalmAndPosters.com, Google, @dan, @teamsteem, @jerrybanfield, @dantheman, wethepeopletshirts.us, harvardlunchclub.com, Youtube
Copyright (part 1 + 2): Seinfeld, Reader's Digest, Jersey Shore