Dilution of quality in social aggregators

I generated two normally distributed samples (10,000 large)for post and comment value. Each of these pairs (or vectors) is considered to instantiate a possible user (PO).

Friend networks for each of these possible users is instantiated by randomly choosing 50 possible users from the set of 10,000 such that the distance between the PO and friend candidates is less than .1 (arbitrarily, but small enough to make a tight cluster of friends).

The initial user set is randomly chosen from the 95% percentile of this 10,000 vector set, and new users are inducted through the networks of these existing users at a small rate.