You are viewing a single comment's thread from:
RE: Improving Steem’s rankings to cater to diverse content preferences
I agree about the imperfections of the issues with the ranking and reward systems. However, I do not see the solution as fully considered in its present state.
Primarily, what are the implications for the allocation of rewards across tags? How much should #steemit get versus #beauty?
Might multiple cluster membership or topic modeling be a better fit?
And by automagically, I assume machine learning of some variety?
Multiple cluster membership is what I alluded to in the last paragraph of the last section:
You are correct to point out I was vague in the last section, but afaics the rewards should still be according to voting power; thus the relative allocation between clusters and hashtags could remain an orthogonal attribute of the voting power. In other words, the total reward for the blog post’s author would still be based on the sum of the voting power for it, but the rankings would be clustered. However, the voter’s reward should be constrained to the voting power of the clusters he/she is a member of, so that the Nash equilibrium of voting for the posts with the highest rewards is I think removed. The proposed algorithm incentivizes the voting power to divvied up more granularly by providing orthogonal rankings for like-minded clusters of voters and this could be computed orthogonally for each hashtag. I will be pondering more the game theory ramifications of this proposal.
I did actually think of this issue, but I just forgot to write it into the last section. My energy level and focus was tailing off apparently as I composed the end of the document. There were many details I thought of and I may not have written them all down yet. I will as I remember.
The automagic is the algorithm as described.