Proposal: reduce Hive inflation by reducing curation rewards

11 days ago
8 Min Read
1532 Words

Background on curation reward problems

The last major changes to the author/curation rewards were made as part of the Economic Improvement Proposal (EIP) changes. The most significant of these changes was a separate downvote mana pool (to allow downvoting that doesn’t economically hurt the downvoter) and a change from 75%/25% to 50%/50% in the rewards ratio paid to a post’s author versus its curators.

Both of these changes were primarily intended to discourage two practices that had become quite common: 1) buying votes for posts from bid-vote bots (these were bots that you could pay to vote for your posts) and 2) self-voting of near content-less posts. Overall, the changes were pretty effective, so much so that vote bots pretty much went out of business.

I think it’s fair to say that curation quality did improve as a result of the EIP. The trending page has better articles and downvoting has curbed a lot of the worst abuses of “bad” whales upvoting crap posts for the author rewards.

But I also think most people would agree that we’re still far from an ideal system for rewarding posts.

The rise of auto-vote bots

The biggest problem I observe nowadays is the rise of auto-vote bots (bots that you give permission to vote for your account). These have existed for a long time, of course, but a lot of the funds that previously were invested in bid-vote bots have moved to auto-vote bots. The problem with this is that auto-vote bots are generally programmed to vote for a set of “known good” authors, instead of voting based on the contents of individual posts. So the judgement of post quality is far from ideal.

If automatic voting is bad for the network, why do large stakeholders do it?

Stakeholders allocate their funds to these bots because they want to receive curation rewards, but they don’t want to spend the time to curate. This might just sound lazy, but it’s more complicated than that.
Curation actually is very time-intensive. So intensive, the rewards just aren’t great enough to pay for the time, so the true manual curator usually does so for other reasons than the rewards. So, if you’re a large stakeholder that wants to hold the token, but doesn’t want to curate, you have four choices: 1) sign up with an auto-vote bot and collect your curation rewards (somewhere between 8-12% APR), 2) randomly curate 10 posts a day and risk those posts getting downvoted for bad curation (losing the curation rewards), 3) devote hours a day to reading many articles and curating manually, and 4) don’t vote and fall behind in Hive ownership versus other stakeholders in the first 3 groups.

Now, while it seems fair that people in group 3 should earn more than people in group 4 (after all people in group two are providing useful work to the blockchain), I think it’s clear that people in group 1 and group 2 shouldn’t earn more than people in group 4. I believe a lot of large stakeholders go through this same thought process, and if they care about an 8% return but don’t have time to curate 10 posts a day, then they decide the only reasonable option is to join an auto-vote bot.

Solution 1: pay manual curators like it’s a job

The problem with this solution is that it is just too expensive. At least right now, it takes most people too long to find a bunch of posts they like. And while large stakeholders get paid more for the same effort and should therefore be more incentivized to curate, they also usually end up being people who are already highly compensated for their time elsewhere (this tends to be just the nature of capitalist systems).

Solution 2: make it easier to find good posts

This is certainly a worthy cause, and we should invest effort in creating better tools for finding good content with less effort. But I don’t see this as a silver bullet solution to this problem right now.

Solution 3: lower curation rewards?!?

After thinking about this for a while, I believe this is probably the simplest near-term solution to the economic problem that is currently pushing curators to use auto-vote bots.

Most stakeholders know that auto-vote bots aren’t great for Hive as a whole. But it’s a form of tragedy-of-the-commons, where each stakeholder who doesn’t join in the auto-vote game loses out against those who do.

To break this negative economic incentive, I propose we lower curation rewards to the point where the potential gain from curation is too small to make it worth the negatives associated with using an auto-vote bot. This will bring the situation back to one where stakeholders who don’t want to spend time curating, but don’t want to lose out on curation rewards, don’t need to worry about this issue any more, because the relative loss between the two choices is small. This could even incentivize new stakeholders who previously have decided to pass on Hive or only maintain a small investment, because they didn’t want to play the curation game.

A side benefit: Solution 3 also lowers overall blockchain inflation

The Hive blockchain currently has an inflation rate of 8.5%, which decreases 0.5% each year (don’t quote me, as maybe it already dropped to 8%, I don’t know exactly when the change occurs). My back-of-the-envelope calculation leads me to believe that overall inflation will drop to about 6.5% if we reduce curation rewards by 75%.

Could this cause the return of bid-vote bots?

One potential concern, of course, is that this change could lead to a return of bid-vote bots and self-voted content-less posts. But the primary disincentives for these behaviors continue to exist from the EIP changes: 1) author rewards were reduced and 2) downvoting such posts became economically net-neutral to the downvoter. So I believe that a reduction of curation rewards won’t result in a return of these behaviors.

Will this reduce incentive to stake Hive?

Yes, I think this change will lower the incentive to stake Hive. The current high APR on curation rewards does act as a strong incentive to powerup Hive. So lowering the curation rewards will certainly lower the incentive to power up. But at the same time, it could act as incentive to hold Hive as a whole, since the more nervous investor can keep his holdings liquid without losing out as much economically versus powered-up stakeholders.

And there will still be good reasons for people to power up their Hive: the ability to influence Hive governance and the reduced inflation impact on powered up Hive (inflation paid to VESTS).

Other consequences: expect less votes

Another aspect of this change is a likely decrease in the number of votes cast for posts. My best guess is that the majority of votes are currently cast by auto-vote bots. So we can expect the number of votes cast to drop quite dramatically with such a change, and to someone who wants to inflate activity metrics, this may be considered a negative.

But I don’t think we should focus on false forms of activity of this sort. Too many blockchains play these kinds of games, but I hope that here on Hive, our focus is on providing useful services for our network participants, not gaming metrics to distort reality.

OK, if you’re still with me on this, how much lower should curation rewards be?

It’s a difficult question to answer, because the answer revolves around human behavior.

Obviously, we could simply drop curation rewards entirely, and I think it’s fairly clear that auto-vote bots would quickly go out of style. Another plus of this approach is that it would also eliminate the need for the blockchain to compute and report curation rewards. On the other hand, this would eliminate one of the few current gamification aspects of the blockchain (competition for curation rewards).

My initial thought is we should cut curation rewards to about ¼ of their current amount. This would bring the reward down to 1-3% APR for most curators. I suspect this would be enough to encourage stakeholders to look at the big picture, and stop delegating their votes to vote bots, allowing manual curation to dominate.

Feedback wanted

All of the above is an attempt to analyze and predict human behavior. That means it’s easily subject to error. I’m not certain that the proposed change will work as intended, it’s just my current opinion on a simple solution to the auto-vote problem. So I’m not wedded to this idea, and I am open to other solutions, as long as they don’t involve a lot of programming work (unless someone else wants to code it cheaply).

In the same vein, I’m OK to leave things as-is, if most stakeholders think that the issue I’m raising isn’t a problem worth worrying about. But all feedback is welcome, as I think it’s an issue at least worth discussing. If there seems to be enough interest in this approach, I can create a pair of proposals to more accurately measure stakeholder opinion about it.