Very interesting and well thought out proposition.
Do you think we could roll something like this out in waves of varying detail?
For instance: coming up with an algorithm that somewhat normalizes the rewards pool into different topics/tags, which then get split up based on the relative success of posts within those topics. I feel like this would be a step in the right direction without also trying to take into account each I dividual user's preferences which are subject to change over time, and other factors like that.
Good point. I forgot to address changing user preferences. To adapt to changing user preferences, the set of content instances considered by my proposed algorithm could be a sliding window excluding old content, or probably better would be to weight the older instances less in the computation of the mean for Dₛₜ.
So far, I can’t think of any computation of rankings by tags without the clustering by voter like-mindedness, which is any different from the current algorithm; and thus won't stop the optimum voting strategy from being for everyone to vote in a groupthink for the most highly voted content.
Thanks to everyone who appreciated the proposal. We need to spend more time thinking about any vulnerabilities it might introduce. I wanted to get the proposal published as soon as possible. I will be pondering it.
yes good proposal