GMs having a mental model they develop and evaluate is only more like our moment-to-moment existence as long as we are dealing with environments that have a fairly definite status quo that player characters are exploring and interrupting. Particularly if they are foreign environments player characters have little meaningful connection to or knowledge of that precedes play. It happens that this aligns very well with the sorts of fictional situations we see in most adventure fiction and the more popular tabletop RPGs.
This sort of breaks down when we move into the sorts of fictional situations where NPCs are as apt to disrupt the status quo as player characters or different characters are interacting with different parts of the environment or there's more familiarity and connections involved.
Take a fairly simple court scene where you have 3 different characters petitioning the king about how they think the would-be assassin of the prince should be handled. One is spymaster who believes the assassin should be cut loose so his agents can track them back to the person who hired them. One is the queen who wants a public execution. One is the master of arms who wants the assassin questioned and then disposed of. 2 of these characters are player characters. 2 are NPCs. This is a confab that could take a day or two with interjections by other NPCs along the way. Playing this out in a way that feels real and gets close to our moment-to-moment experiences is much, much harder to handle in a GM's mental model sort of way. My experience is that at least some active scene framing helps to bring this sort of situation to life.
This is why I think a focus on how explorable the situation (and what level of zoom we are operating at) is more apt than trying to argue over which approach is more real (which is going to be very subjective and contextual to the sorts of fiction we are dealing with).
One of the things that is always bizarre to me is that our discussion of different techniques and approaches always seems to land on the sorts of fictional situations that are most appropriate map and key task resolution. Locked doors and mountains to climb with sparsely detailed fictional situations that give them none of the rich situational context* that other approaches require to functionally address.
This is not to say that a rich situational context is not amenable or desirable with map and key task-resolution oriented play. It's just like not necessary for functional play or addressing the GM decision space in the same way it is necessary in other approaches.
Right, so to kind of give some historical perspective: you can look at the evolution of RPG design and practice as a process of both making the GM's role easier, or at least more practical. Coupled with this is a broadening of the kinds of situations which these games can handle.
Early proto-RPG play, such as Braunstein, required a lot from the GM. Essentially it was necessary to devise a whole set of heuristics and rules, props, etc. and then manage play in a very unstructured way. Note how only a bare handful of people have ever put these things on. Even if we include modern who-done-its and such this is a tricky thing to carry off and requires substantial planning (I think
@pemerton has mentioned putting one on).
Dave clearly understood what
@Campbell is talking about. He utilized wargame-based structure to impose better roles, conflict resolution processes for, at least, combat, etc. But note the result; D&D was entirely focused on this very stylized exploratory context. The other TSR early RPG, Boot Hill, was based on a wargamization of the Brownstone TX 'Braunstein', consisting of a square ruled town map and rules which covered various forms of combat, riding, and gambling, bar fights, stage coaches, etc
Note that, even with d20 style skills, none of this easily applies to unstructured situations. It would require the development of intent-focused for that. Once you have those, you can then address situations like a negotiation or puzzle solving in a game structure.
And that's where other additional elements come into play, generalized scene framing and system support for discerning intent.
