I like the idea.
Humans would also quickly glance at the type of commenting, and detect fake, superficial comments like "good job! nice post!".
I make comments like that quite often and they are neither fake nor superficial. Sometimes you want to tell someone you enjoyed the post and you don't really have anything further to add. That doesn't mean the comment is any less valid.
what i meant by that is, if the system is trying to be gamed with a lot of short, fake comments that are just generated by either bots or armies of people trying to create the appearance of engagement.
I'm sure there would be a way to test if a bot is just repeatedly making the same comment using some kind of automated analysis - I doubt they vary what they say.
sure, there's probably a way...
I'm sure it can be done.