I've been working through Sine Nomine's Labyrinth Lord setting, Red Tide, and how it presents an unorthodox approach to structuring a sandbox. In Red Tide, it's suggested the DM builds some prebuilt content to deploy in the sandbox when the group moves into a new area. It's compelling to me because it purports to provide a way of cheating the sandbox triangle (in the absence of unlimited prep time, there are trade offs between detail and scope).
The problem, of course, is whether dropping prebuilt stuff in front of the players is foul play? The discussion - here at my place - The Shell Game in the Sandbox - and over at Hack & Slash - How an Illusion Can Ruin Your Game - has focused a lot on a slippery concept we refer to as "agency" - do the players have freedom of choice, and are their choices meaningful?
The problem I keep coming back to, when determining whether the DM is running a game that supports player agency, is that player agency is subjective. On the player's side of the screen, they don't know if the DM is being a malefactor; their ability to judge is based on the information they have and whether the world is behaving within expectations.
There is a more objective standard that can be applied to the problem of agency, and it's similar to the definition of railroading - is the DM predetermining the game's outcomes? It's an objective measure when viewed from the DM's perspective. We can say with certainty that when the DM is predetermining the outcomes, player choice is rendered meaningless because the DM is railroading, even if the players don't know it.
So using the railroad language gives a simple test to determine whether using the Quantum Ogre is a bad thing, and if there are circumstances when it's okay to employ it. I'll start by revisiting the shell game.
The Shell Game
The shell game goes like this - the party is looking for a MacGuffin in the woods - it could be in woods A, B and C. No matter what woods the players enter, the DM puts the MacGuffin into the last wood. We'd all agree this DM is being a bad actor. It's harder to frame an argument that he's not supporting player agency. But objectively, the DM has predetermined the outcome; playing hide the MacGuffin is a railroad.
The Quantum Ogre
The problem of the quantum ogre is this: assuming the party has no additional information, there may or may not be an ogre in the woods. Once again, he's not necessarily violating agency from the player's perspective when the ogre is dropped in front of them; the player choice was to enter the woods and search for the MacGuffin, and it's a reasonable expectation there could be an ogre in there. From their viewpoint, there's no difference between these scenarios: walking into the woods and running into a scripted ogre they failed to learn about; a randomly generated ogre via an encounter table; or the ogre the DM dropped on them through sheer fiat.
But when the DM *always* places his ogre encounter in front of the players, he's predetermining an outcome and perpetrating a railroad. We can always identify the bad behavior from the DM's side of the screen, even if the current state of player knowledge doesn't allow them to see it.
In Defense of Dice
Over on Hack & Slash, I pointed out that there are numerous ways to stock a dungeon or wilderness area - a spectrum that extends from pre-stocking everything (no variable encounters), to mixing pre-stocked with random, to purely random, to purely improvised.
It's part of the agreed foundation of the game (the social contract, as the cool kids say) that certain game results will be determined by dice. Dice are a neutral arbiter; it's why never fudging the dice is such a powerful maxim. Dice protect the DM from himself. The dice also provide an answer to the problem of the quantum ogre.
I know this seems really self-evident - it's why we've used random encounters in this hobby for so long. They're inherited wisdom. The random encounter probabilities provide a way for the DM to introduce a degree of variation in the game results at the table, in a matter agreed upon by all participants, while limiting the DM from pushing predetermined outcomes (ie, there's a thin line between improvisation, making shit up, and outright elimination of player choice).
In the case of the DM's cool ogre encounter, incorporating the ogre encounter for a result on a random encounter table is a way to seed the encounter in the setting at run-time fairly, by relying on the dice as neutral arbiters. The side benefit is that this typically ends up being a better table experience than a scripted game. It's also the reason I encourage reaction rolls and morale rolls to help determine how NPCs respond to player actions; the DM is within his rights to mandate NPC reactions, but keeping the door open for an unexpected result via the dice ensures that predetermined outcomes are avoided.