This is a really cool insight but I've been sitting here trying to formulate a response for a while, and like. This sounds like a really good model for why and how to do things; in terms of being-able-to-predict-yourself it might even predict me better than the concious models I use for deciding how to follow my own preferences (which tend to be world-state-optimization and/or resource-optimization shaped)

"Why do people care about things" seems to be the hardest and the easiest question sometimes, but I guess it's not actually that useful to answer unless you're distinguishing first-order wants from nth-order derivative wants.