This may be true, but this complexity gap need not stop us any longer. Why bother simulating any social or emotional nuances when real live humans are so easy to co-opt into willing participants in a social space facilitated by web technology? Don't waste your time writing advanced AI, just use people.
Say you have some item, like a virtual item in a game. If you are making a social game, you might be tempted to try directly specifying its effects on a player's social network. This would be hard to do effectively. But if instead you structure your players' social interaction and communication to meaningfully involve and be affected by this virtual item, these players will construct a much more nuanced meaning for it themselves. There is no need to abstract it away.
Provide opportunities for the players to create significance themselves out of the elements you provide. Come to think of it, this approach can even apply outside of social games, sort of like how Tale of Tales would like to see more games as instruments.
Now I think that abstracted formal simulations of social dynamics have their place, particularly for educational purposes or to illustrate a certain point. But I wonder if either of these goals might still be compatible with the structured socialization approach, using real players. It's an interesting idea. How far can you go in simplifying these interactions and dynamics before people are unable or unwilling to play in your world?
*image from the excessively cute and catchy Less Than Three*