With July 4th rapidly approaching we are running out of time to review and tweak how curation rewards get applied. One of the reasons we opted to wait until July 4th before paying out rewards was to give us time to evaluate and tweak the algorithm based upon actual usage patterns.
Lessons Learned
Over the past month and a half we have learned quite a bit about user behavior and interaction with Steem. Here is what we have learned:
Curation Rewards don’t help the little guy
The exiting curation reward allocation strategy heavily biases rewards toward those with a large stake. This creates a centralizing effect that doesn’t sit well with many people and ultimately defeats the goal of making it easy to recognize participation.
Bandwagon Effect
The existing algorithm encourages people to pile on a single post. This distorts voting behavior and ultimately could result in unintended / biased curation results.
Big Bad Bots
There are bots that get the jump on humans. While many people correctly point out that bots add value, there is a very real feeling that they suck up a disproportional amount of curation rewards. This combined with bandwagon effect could evolve into undesirable behavior.
Cognitive Load
People face a cognitive load trying to figure out how and when to vote. This makes voting stressful and causes people to vote less than they otherwise would. We would like to reward honest people voting honestly without making people play a game. We are successful if honest behavior is rewarded more than attempted abuse. We fail if people spend more time thinking about their rewards than they do actually evaluating content and expressing their honest opinion.
Comments get little Love
One of our original goals was encourage quality discussion. The algorithm as it exists now discourages people from voting on comments. This happens because it dilutes their voting power which in turn reduces their potential curation rewards. Those posting comments get significantly less votes than those posting top level articles which means they often get next to nothing for their effort.
A new Approach
Our team has been discussing these problems and potential solutions at length. Many on Steemit have also proposed interesting ideas. Ultimately we concluded curation rewards would have to take a radically different approach if we wanted to maintain a more egalitarian platform. So we started asking ourselves what kind of behavior do we really want to see. This is what we came up with.
Vote or Post Every Day
Users who login and do something every day remain engaged. This engagement will naturally lead them to discover new content they will want to vote on. We don’t care what they vote on, so long as they show up every day.
This can obviously be automated by bots, but it has the important quality that bots have no advantage over regular people. The network already rewards “bots” who produce blocks and mine so giving them a small extra bit for voting daily doesn’t really change much. It is more important who isn’t getting the rewards: those who do nothing.
Discuss Quality Content
If a post is going to earn $4000 of rewards, then the community had better discuss it throughly. Any content that isn’t worth discussing is probably not worth rewarding. It is the discussions that ultimately provide ‘reviews’ of long articles. Discussions are where fact checking occurs. Plagiarism is also identified in the discussion. You could easily argue that the discussion is a better kind of curation than voting. Discussion helps people decide what to vote on and is ultimately something that bots cannot do.
More importantly, it is far easier for most people to post a quick comment than to write an article that becomes popular. This creates more opportunities for average people to be recognized on Steem.
Producing Quality Content
It has become clear that Steem is evolving toward rewarding original content over rewarding cross-posting content. We need to maximize the incentives for original content creators above all.
Proposed Allocation of Curation Rewards
Here is the new algorithm that we would like to use to divide / allocate the curation rewards:
Activity Rewards
30% of existing curation rewards will be redirected toward activity rewards. This represents a total of 15% of all (content+curation) rewards. To maximize your activity rewards you must vote, post, or comment at least once every 24 hours. These rewards are distributed proportional to the amount of Steem Power your account has. There is no advantage to being more active other than the joy of reading and rewarding your favorite contributors.
To minimize abuse and/or delegation to bots, only accounts with simple posting authorities consisting of a single key can qualify for activity rewards. If you attempt to use advanced multi-sig configuration on your posting authority then you will not quality for activity rewards.
Comment Rewards
50% of existing curation rewards on top-level posts will be redirected toward comments on that post. This means if you see a post with a pending payout of $10,000 that $2,500 will be allocated toward comments on that post. Comments at any level will divide up the $2,500 proportional to the total amount of Steem Power voting on the comment.
Removing Parent Rewards and increasing Author Rewards
We originally intended to reward the parent for stimulating discussion by sharing part of each comment’s reward with their parent. This process has been removed because it contradicts the direction of payments that we now have for comment rewards. To make up for lost parent rewards, the remaining 20% of existing curation rewards will be given to the author.
Summary
Steem currently allocates 2 STEEM every block toward content + curation rewards, after these change the 2 STEEM will be divided like so:
60% to author of post
25% to comments (at all levels) on original posts
15% activity rewards
Conclusion
Under the new rules it should be possible for more people to earn Steem Power just by being involved each day and adding to the discussion. Content creators will have an active and engaged audience who will likely share what they are discussing with others via other social media platforms. We will have greater ‘curation’ as people post comments that review / critique / add to the original content produced.
People will have financial incentive to be first to reply to potentially popular content with a comment that itself will likely receive many votes. This process makes the people voting a turing test for curation and minimized the potential profits of voting bots.
Overall we feel that these changes should maximize the community participation.