the rating system could be circular (by lack of beter name)
users would up and down vote things they like or not. value is them extracted from contradiction(!) if user A liked content X while B does not and user B liked Y we assume A does not like Y. with lots of fine gradients like that I assume the high noise will eventually be drowned out by the signal.
users would up and down vote things they like or not. value is them extracted from contradiction(!) if user A liked content X while B does not and user B liked Y we assume A does not like Y. with lots of fine gradients like that I assume the high noise will eventually be drowned out by the signal.