I'm trying to learn more about trust metrics (including related algorithms) and how user voting, ranking and rating systems can be wired to stiffle abuse. I've read abstract articles and papers describing trust metrics but haven't seen any actual implementations. My goal is to create a system that allows users to vote on other users and the content of other users and with those votes and related meta-data, determine if those votes can be applied to a users level or popularity.

Have you used or seen some sort of trust system within a social graph? How did it work and what were its areas of strength and weaknesses?