Maybe a similarly crude level of simulation could work for social games, if the artwork allows players to get immersed in the game and fill in details with their imagination. Japanese relationship games and dating sims get part of the way there; since they're "visual novels," they can convey subtleties of character through narration, instead of having to come up with game mechanics for everything. The actual mechanics are little more than a single love/hate score for each girl, with the player successfully dating one when their score gets high enough (and getting a "bad end" if all are low -- or if too many are high).
Chris Crawford on Game Design mentioned that for the game Gossip, he simulated relationships as a network of springs. (http://en.wikipedia.org/wiki/Gossip_%28computer_game%29) The closer Alice is to Bob, the more she can influence his opinion of (distance from) Carol. Another spring-based AI by Michael Heasell (Wizard Dude) won CodeCombat's programming tournament. (http://blog.codecombat.com/a-31-trillion-390-billion-stateme...) This "system of springs" could be the core of a social game the way "collision detection" is the core of shooters.
I suspect that one could create a relationship simulator with an emergent storyline by modeling the characters' relations towards each other (not just having a love/hate counter for their relations with the protagonist), and by having a skilled writer create dramatic scenarios triggered on certain conditions and "everyday" scenarios to move the plot along and provide chances to adjust the relationships. The standard plot tree of a dating sim would become a plot graph. It's still a far cry from real human interaction, but Call of Duty is far from real warfare and it sells anyway. The hard part is modeling just the interesting parts and making an entire playthrough be "fun."