I am not talking about "input from the meta level". I am talking about rules that govern the GM's decision-making. Just as you are, when you refer to "a consistent world", which is (as best I can tell) a heuristic you advocate GM's use to make decisions about what happens next in the fiction.
I am only talking about the first of these.
I am talking about RPGs like Burning Wheel, Torchbearer and 4e D&D where the GM has authority to frame scenes.
As far as determining what is at stake and influencing tone and theme - if you are saying that, in a living world sandbox, only the GM can do this, then to me that only drives home how great is the GM control that you are advocating for. Given that stakes, tone and theme all follow from GM decisions about their PCs, what they care about, and what they do, as far as I can tell these all sit within "character agency".
No it wouldn't. As best I understand it, Arneson was not running Blackmoor in a manner that made it hard for players to understand what the consequences of their action declarations would be.
You say you’re only talking about character agency, but the systems you reference, Burning Wheel, Torchbearer, and 4e D&D, rely on structured procedures that constrain the referee. These games give players control that extends beyond just acting as their characters. They also allow players to act for their characters, meaning they make decisions that affect the fiction at a structural level, such as framing scenes or establishing stakes. That is a different kind of agency than simply declaring actions from within the character’s point of view.
This distinction is important. In a Living World sandbox, players act as their characters. They make decisions based on what their characters see, know, and understand, and those decisions have consequences grounded in the logic of the world. The player is not stepping outside the fiction to shape the narrative or declare what the session is about. They are not authoring outcomes or negotiating framing. That kind of input is acting for the character, and it belongs at a different level of play.
You claim that stakes, tone, and theme follow from what players do with their characters. I agree, but in the Living World approach, those elements are not prepackaged or negotiated in advance. They emerge through the ongoing impact of events and consequences. Players are not told what the stakes are. They find out through play. Tone comes from how the world responds, not from a shared declaration. Theme arises after the fact. You only seem to recognize player impact when the system gives players structural control. I recognize impact when it emerges from consistent decision-making within the world.
As for Blackmoor, you say it was not railroady because players could predict the consequences of their actions. Yet Blackmoor lacked formal procedures for resolution. What it had was a referee, Dave Arneson, applying judgment based on world logic, for him it was history and what he loved about science fiction, horror and fantasy. That is the same thing the Living World model does. If Blackmoor had agency, then Living World campaigns do as well.
The real issue here is that your usage of character agency folds procedural guarantees into the definition. You say you are not talking about meta-level input, but you only seem to acknowledge agency when the system provides formal control mechanisms. That treats acting for the character as the default model, rather than recognizing the legitimacy of acting as the character in a consistent world.
It is OK to prefer that model, but it does not make other approaches flawed. It means they are working from a different definition of agency, one that deserves the same consideration.
