A new approach to Content Reward Allocation Part 2

I would like to thank everyone who responded to my prior article, A new approach to Content Reward Allocation. I have had some time to think about the responses and would like to address the concerns here.

Complexity

I hate complexity and desire to minimize it. Very complex systems can hide behind very simple user interfaces. Just think of the complexity behind your cell phone screen. Let's not fear complexity if it can be properly wrapped in a simple user experience.

How a Bot Might Vote

I put myself in the shoes of a bot operator attempting to maximize my profits from automated voting. If everyone is voting on a scale from 0 to 5 then I would always vote 3 to minimize my potential for error. As a voter I do not make more money by voting 5 or 0 because my reward (and the reward of other voters) is not tied to the score I give the post. The only person impacted by the score is the author.

Knowing this fact, anyone playing the odds will vote the global average of historical votes. A slightly smarter bot would use statistics about existing voters history to infer how they likely voted. In any event, as a bot operator my goal is to "vote the average". I will know that all other bot operators will be doing the same. This means deviations from 3 will only happen when people vote for reasons other than curation rewards.

Ease of Use

Most people expressed concern over the fact that the proposed algorithm would complicate the "ease of use" for the website. I think it is fairly simple to make the default upvote 3 and the default downvote 0. Downvotes are not rewarded, but then again people don't downvote to get rewards, they downvote to punish abusers. A slightly more advanced interface would use the same algorithm as a bot to predict the vote and apply that.

The impact of a downvote is to move the average. In this case a few whales downvoting could have a dramatic impact on the average. This will have a dramatic impact on the rewards of bots who always vote 3 or near the average. Those bots who are slightly more sophisticated will make more dramatically money by being slightly more accurate than those using the dumb averages.

The outcome of this is that bots profit by predicting the ratio of upvotes to downvotes.

Optional Advanced Interface

Without changing the blockchain, the user interface could enable a 5 or 10 star system that would give human voters slightly more flexibility in assigning value. These advanced users may simply want greater control over whether memes make more than investigative journalism. An advanced user is looking more toward the long-term value of their Steem Power than they are for short-term curation rewards. When they vote something to be a "5" they are simultaneously rewarding the author and punishing those who use less sophisticated curation methods (by increasing their prediction error).

Outcome

The predicted outcome of this will be a strong bias toward the average with profits going to those who more effectively predict deviations from the average. The user experience of Steemit wouldn't change from the simple interface we have today.

Rate Limiting Voting

This new system would have to rate limit voting much like the current system does. The impact of rate limiting is to create an opportunity cost. The most curation rewards are given to the posts with the highest quorum so using your bot to vote for one post comes at the opportunity cost of voting for another post. In effect, a bot operator makes the most money by predicting the most popular average content. Popularity is measured by the number of users who engage with voting (up or down) a piece of content.

Removing Blinding

The algorithm can work even without blinding. The dumb bots would just vote the current average. Users make the most by predicting the future result. Once you remove blinding the new algorithm and its simple interface would still work well. The impact of removing blinding will be to heavily bias the final result with bots reinforcing the initial predictions of users.