So the more the GM controls of all this - where, as I've posted upthread (making the analogy to other games, like chess and bridge), control is not the same as authority - the more of a railroad it is: the GM is controlling the shared fiction (perhaps in response to player prompts). The converse of this is player agency - the players exercising control over the shared fiction. The asymmetric roles in the game mean that the GM will always have some control, particularly via framing and some aspects of outcome/consequence. But when the GM has all of it, or most of it - eg all the players are doing is to declare what actions their PCs take, but all the rest is with the GM - then that's what I call a railroad.
Reflecting on the issue you raised regarding player agency, referee control, and the shared fiction, it felt familiar. Then it hit me: it raises the same concerns that Ron Edwards addresses in
his essay "Narrativism: Story Now."
What Edwards calls The Impossible Thing Before Breakfast is the contradiction that arises when a game assumes the referee is the sole author of the story, while players are also supposed to direct the actions of the protagonists. If the referee controls the story’s direction, then player choice becomes cosmetic. If player decisions matter, then the referee can’t pre-author the narrative. You can’t have both, at least, not without undercutting someone’s agency. That’s the core tension.
I see that your approach differs from Edwards’ in that it's more practical than theoretical. You’re focused on the procedures at the table, how control over the shared fiction is distributed through framing, resolution, and consequence, not just on abstract authorship. You're tracking agency through what happens in play, not through philosophical premises.
My living world sandbox avoids the contradiction Edwards raises because I’m not trying to tell a story or construct a narrative arc. I’m focused on presenting a setting that operates independently of the players, a world in motion. My goal is to bring that setting to life so that the players feel like they’ve truly been there as their characters, making meaningful decisions and pursuing the adventures they choose. If a story emerges, it does so after the fact, as an account of what happened, never because I authored it in advance.
Likewise, my campaigns don’t engage with the contradiction you’re pointing to, because we don’t treat the shared fiction as something co-authored by the group. The players understand that they are visiting a world that exists outside of them. The only way they affect it is through their characters’ actions. What changes is a result of play, not through any negotiated or shared control of the fiction.
Because of that, my campaigns aren’t railroads in the classical sense, where player choices are blocked or overridden. And with respect to your definition, where railroading is about the GM controlling all or most of the fiction, it’s simply not a concern for us. Neither I nor my players are trying to collaboratively build a story or manage narrative control. Like with Edwards’ Impossible Thing, the premise doesn’t apply. The players are focused on experiencing the world as it is, and testing how much they can change it through what their characters actually do.
Your concern about control over the shared fiction makes sense given the creative goals you’ve laid out. But it doesn’t apply to the kind of experience a Living World Sandbox is designed to deliver. In fact, anything that gives players narrative control beyond their characters' abilities tends to undermine the experience they’re seeking. This isn’t just theory, my players and I tried Whimsy Cards back in the '90s, and we’ve explored games like Fate and other systems that shift narrative authority. They consistently pulled us out of the feeling of “being there,” and that immersion is central to what my players are after.
So circling back, your points make perfect sense given what you’re trying to do with your campaigns. My approach doesn’t engage with those questions, not because they’re invalid, but because they’re not relevant to the creative goals I’m pursuing. That doesn’t diminish your position or Edwards’. It simply underscores that we’re solving different problems in pursuit of different play experiences.