So, this is really early, as in I was just out on my walk and thinking about the curation algorithm being broken, and had an idea, and I want to write it down and get some feedback on it. I totally have not thought this through to proposal status yet, but I think it might be a way toward solving a few of the problems with curation rewards.
Curation was set up to encourage discovering and rewarding undervalued content, but I think by now we're all pretty clear that's not what it does. One of the reasons for this is that the goal is a pretty hard one to accomplish, and this thought isn't going to get us all the way there, but I think it might take a step in that direction.
What if the curation exponent varied with the poster's lifetime rewards?
Right now the curation algorithm runs the same no matter what post it's on, and there are some square roots in it to make it very strongly favor being one of the first people to vote on a post. You can see why the designers did things that way, because the idea that being the first one to find content should be rewarded more is central to the idea of discovery that they wanted.
But in practice, posters tend to put out pretty consistent levels of content, so it became pretty easy to predict which posters were going to get high rewards, vote on them quickly, and get a higher curation reward. People automating that is what got us the awful-and-about-to-get-worse workaround of the early voting penalty.
So as I've been thinking about that I've been thinking about how to reduce the reward for being first to a post that everyone knows will be popular, and increase the reward for being first to a post that will surprisingly become popular, to support the original goal. And most of my first ideas sucked, and I just rejected them out of hand, because this is legitimately a hard problem. And this one might suck, too, but at least it's going to take some thought to be sure of it, which is a step in the right direction.
What I want to do is make the curation reward more linear the more established the author is. Essentially, once someone has been here for a while, and established themselves as a consistently-rewarded poster, the reward for being first to them should decrease. The total curation reward for the post would stay the same, but it would be distributed with less focus on the time of the vote the more established in the community the poster became.
The posts of consistently top-rewarded posters would have a curation algorithm that was linear, essentially as DTube does theirs now; you'd get 25% back for voting on them, period. Meanwhile we'd keep the n2 algorithm for new posters and posters who haven't seen very many rewards, so being early to an unexpected quality post would still offer a substantial benefit. In between them the exponent would scale, so there would continue to be incentive to vote on someone early as they grew, but the more established and connected they became, the more regular votes they earned, the less that incentive would be.
Advantages as I see them:
- This would make the curation system more effective at discovering new content.
- This solves many of the problems the early voting penalty was meant to address, so maybe we could scrap it and all its nasty side effects.
- If it works it might be a way to compromise between the two sides on increasing curation rewards percentage overall. I know I'd be more willing to see that happen with a curation system that worked better.
- It makes bidbots and vote-buyers a little less disproportionately powerful. The bots mostly wouldn't take more than their share of curation rewards anymore.
Potential flaws I've noticed:
- It doesn't solve all the problems the early voting penalty was meant to address, especially a potential workaround by starting a new account and immediately botting it up a lot.
- It might reward periodically starting new accounts in general. I think regular voting support would be worth more than a favorable curation algorithm for the vast majority of posters, but there might be a population I haven't thought about.
There are probably more flaws I haven't noticed, so tell me about them in the comments.
