"in Oblivion, each NPC exhibited goal-oriented behavior to better mimic what real-world people might do, sometimes leading to amusing and unrealistic scenarios."
Yes, but what does that lead to in terms of teaching? Is one going to go from autism to a deeper understanding of human motivations by playing Skyrim? No. It isn't anywhere near rich enough to do that, and it isn't because we wouldn't like to have characters in a game that rich, it's because we really don't know how to make them. It's all skin deep. Pretty much everything is.
Since my thesis is that we don't know how to make highly developed interactive environments that one can learn from, I am at a loss as to why you think I'm going to come up with an example of one.
We've invested an enormous amount of money into making things look good. There's nothing wrong with that. But we have, as a consequence, the ability to make something that looks almost photorealistic and looks like the real world, where in fact it's just a glorified toy, and it fools people into thinking we have a lot more power to create games than we actually do. At best we can manage tinkertoys or legos, as in the Civilization or SimCity examples I gave. But how would we build a game to teach History qua teaching History? To actually teach it and not merely allude to it, or merely wrap the game up in the stylistic ambiance? Or merely be a glorified flash-card simulator? Beats me, and I've been pondering the question and watching everybody else's attempts to answer the question for a long time. I don't see any evidence anyone has anything like an implementable idea at our current level of technology.