Seems to me the missing element from your analysis is simple: Users don't need to have exactly the same experience. Picture a graph where each user is a node. W

Seems to me the missing element from your analysis is simple: Users don't need to have exactly the same experience. Picture a graph where each user is a node. When I freshly arrive at the site, it tells me to pick out a few users to place in my "circle", to use Google terminology. I do this by finding users who write comments I think are good, or by looking at a list of some of the most popular users. I can then set the degrees of Kevin-Bacon controller to set expand my "circles" outward by joining in people that others in my circle trust, multiplied by a reduction factor (farther out means less influence). Only votes by members of my extended circles count towards my "score" for a given item, weighted by distance. If something is upvoted in my perspective, I can see who and why, and break that connection (remove a person from my circles). Finally, I can publish a group of users I trust and give it a name, which makes it easy for others to follow well-known groups. I think it would work. A user has to put in a little effort, to find people they think have good judgement. It's very easy to knock out bad judgments by breaking links. If I were less tired I'd think this through. The downside to all of this is processing time, of course.