In search for perfect curation system

5년 전

Couple days ago I read @jerrybanfield's disturbing post about current best curation rewards strategy. In short, he concludes that using an upvoter bot is far superior to any attempt to vote by hand.

Can't cope with injustice!

This picture (and article!) summarizes my rant about this sad fact. Like many people in comments to Jerry's article, not only I find situation discouraging, I see it as real threat to the platform. Unfortunately, it seems to be an inevitable consequence of how curation rewards are distributed.

Minnow curation is hard and underpaid

Anyone who tried to manually read through feed of new posts knows how rarely gems that worth an upvote are found. This is not a joyful experience at all, I should say. It is hard labour.

What is worse is that those few who volunteer for this work, quickly learn that it is a Sisyphean task. Even when they manage to find a hidden gem of unpopular author, do upvote and resteem it with all their might, their votes do not bring them any payoff simply because nobody else upvotes the same post. Effort of a few is not enough to grant visibility to an article even when article's worth it.


Popular authors are easy curation money

As @jerrybanfield amply shows in his article,

In summary, upvoting posts from popular authors early with an autovoter consistently brings upvoters 3 to 5 times more than upvoting authors not consistently earning $100 a post. The maximum rewards come from upvoting popular authors that have a high percentage of manual votes

This happens because curation rewards depend on 1) how quickly curator upvotes a post, 2) how much SP worth of upvotes the post collects afterwards. Popular authors are a safe bet in terms of (2), and upvoter bot solves the (1).

Upvoting a good post from a good author is nothing wrong by itself. It's just that bots leech rewards while bringing nothing useful for a community. Popular authors already have great visibility, there's no reason to reward people for "finding" their content. It's reward pool inefficiently spent.


Wrong incentives

For better or worse, people tend to do what personally benefits them most. Hard to blame them! This is the core design principle of any distributed p2p system. Even the hyped blockchain security is based on this idea.

As long as it is safe, benefitial and trivially easy to upvote handful of popular authors, people will use upvoting bots to do that. This twists incentives for popular authors. Why bother writing a good article, when army of bots upvote without reading. This tempts even the best of authors to occasionally get sloppy and post lazy nonsense. And it pays!

And on the flip side, as long as manual upvoting is hard, time-consuming and unprofitable, people won't do manual curation. No manual curation means it's hard to get audience for new authors, even when content is competitive. No visibility for new good authors again lowers quality of content on the platform. Even worse, new authors are tempted to increase visibility of their posts by shady and crooked means like promotion bots, upvote farms, etc. It should not be a norm for comments section to be dominated by a bunch of paid upvote bots boasting about their service, even if the article itself is totally worth upvoting.

All this is wrong on many levels and should not be like that. Sadness.

Can it be improved?

I don't know. Two days of thinking were not enough to get an enlightenment miracle. Steemit community developed pretty freaking good curation system — that is: best I've ever seen!

I came up with a couple of thoughts, though.

We need better tools for manual curation

The "curation is hard" problem can and should be addressed with better tools. Steem community has great developers. Ethereum turns green with envy!

Something as simple as posts feed viewer with filtering by tag, post length, presence of at least one image, filter by keywords, author reputation and author average post payout — this should work wonders for manual curation process. Please make this!

Tools need maintainance

I managed to find some tools that come pretty close.

Understeemed by @tabris comes closest to how I picture perfect curation tool. It has filters by tag (and can show all posts too), number of votes, payout and post length. Post length selector seems to be honed for Korean and asks for larger numbers. It also lacks filter by author reputation (which it does not even show) and average payout. Filter by existence of image or keywords in post would also be nice.

Autosteem by @unipsycho looks impressive, too. But alas, it lacks filtering even more than Understeemed. And it especially scares me to enter passwords on 3rd-party websites. Integration with trusted auth provider is very much required.

Seeress project by @triddin is a Discord bot that notifies about posts that contain certain keywords. If only it could filter by post length and other criteria! What a great curation tool it would be.

For the sake of completeness, SteemNow by @penguinpablo allows to filter Feed/New/Hot/Trending list of posts by author reputation and current post payout. It does not seem to be intented as a curation tool, though.

If any of these tools gets features I dream of, it should revolutionize the Steemit curation. Please help if you can!

We should encourage social curation

Curation should not be about clicking upvote button and moving on. Any bot can do that.

This is not an obvious thing, but what I see good curators doing is this: they generously bring audience to good posts of others. And it pays. They get followers because curator himself becomes a valuable gateway to good content. They also get plenty of curation rewards — and this makes sense! Curator upvotes, then brings in others who hopefully upvote a good post, too. Hence the initial curator's upvote gets a good share.

@curie has a great system of social curation. Check it out if you haven't! Description of their structure was pretty hard for me to find, the best place I know of is the list of pinned messages in their chat channel (pin icon on the right).

In short, they are a group of curators with a set of guidelines and a rating to encourage high standards of acceptance. Entry-level curator upvotes a post and submits for higher-level guy. When accepted, the post receives a full 100% upvote, which currently amounts to something like $86, which is a respectable bump. They intentionally maintain low acceptance rate to ensure 100% vote power on each article promoted. Needless to say, initial curator, hopefully being one of early upvoters, receives a nice chunk of reward.

Humans go social! Don't be a bot!

I believe, something alike the Curie structure should involve the whole community. A hierarchical social system everyone knows about; with lower-SP people promoting good content to increasingly more powerful fish, up to whales — this is how I envision future Steem curation. Maybe I even enthralled you enough to make it eventually happen :)

Thanks for reading!

Authors get paid when people like you upvote their post.
If you enjoyed what you read here, create your account today and start earning FREE STEEM!
Sort Order:  trending

We have to improve our system.

Thanks for your help pointing out the lack of features and improving UNDER_STEEMED :)


Thank you for making it in the first place! :)

I have noticed that some of my posts gets a number of votes that don't match with the number of views, i am not the best writer but some post i think are good enough, it would be great that those posts be seen and read rather than upvoted without cura

I totally agree with this post @teo-nyx. This type of behavior is definitely discouraging in the platform. Makes me sad that some people thinks this way

Could give more to those who bring in the most new readers.
And spread out the curation to a tie for a larger group of people for the best sellers so that more people get tied for highest curation points when voting for a proved popular writer.
Could also give less curation credit for proven writers in form of a lower % since the vote isn’t revealing anything not already known....