something like correctness^2 x novel information content rank?
Actually now thinking about it, incorrect information has negative value so the metric should probably reflect that.
Actually now thinking about it, incorrect information has negative value so the metric should probably reflect that.