I've been thinking about this for a while. What it needs is shareable trust networks, so you can see ratings weighted by the votes of those who you particularly approve of or despise. Attempts at objective moderation just end up creating echo chambers and/or popularity contests.
I have discovered a truly marvelous proof of this, which this thread is too small to contain...but I am making notes to write it up in some useful form.