The outlier model would be interesting, but also potentially irrelevant on a stake based system (perhaps). For examplem if you have a million Steem and look for content that will increase the value of your Steem directly (whatever that might be), there might be a host of people with 10 Steem who just don't care and would rather rank a meme a 10 than what actually could add value to the platform.
In some ways, what we are trying to avoid is the making of the Homer car for content:
I think this happens through niche content markets where the truly interested amateurs and professionals can rank content with a discerning eye, rather than the every man and his dog approach having a say on all content.
I don't go to my hairdresser for heart surgery, nor a surgeon to paint my portrait. When it comes to content evaluation, while we can all be critical, it doesn't mean we have the actual knowledge to critique.
Side note: I cut my own hair.
Good point about niches; I'm already a fan of SCOT tribes because they are helping in terms of creating interest groups and making content discovery a little easier.
I'll be interesting to see how SMTs/Hives impact the Steemosphere; I'm hoping a healthy level of granularity ensues... not too "generic" and not so specific that we end up with a bunch of semi-stale tribes with 7 members, because they are too tightly defined. I suppose we can but hope that the user base the natural intelligence needed to figure it out.
=^..^=
I think there will be communities that could have many thousands, others that have very few members in a very tight niche. As long as they are valuable to the users, it doesn't matter much.