What about an experiment? PG could take a few dozen people he wants to be exemplars for the community, pick some semi-popular articles before and after the change, and ask the test subjects to rate which article had better comments. For best results, only pair people with articles they haven't read.

Myself, I didn't even realize how strong the group-think effect was until the comments were turned off. It feels like there's more substance now, but I might just be reading more closely in the absence of loud communal signals.