I don’t exactly know where I got this idea, but I think it may be cool. On the other hand, it may suck. So I’m asking for your opinion. …The idea is a form of two-player interactive fiction, where one player controls a character as normal, and the other player is a narrator. The narrator is, in terms of the effect on the game, similar to an interpreter, with the obvious difference being that the narrator can respond much more freely. More specifically, there WOULD be a standard interpreter, but its output goes through the narrator, who can change it. The narrator can also give commands to the game world, like “spawn a kitten at the location and make it edible”, using a “scripting language” (in quotes because the “scripts” will usually be one line long). Properties would be defined as needed - in the example, the kitten’s description would not need to be defined until the player examines the kitten.
It’s not really a new idea at all. Back in the days before it was taken for granted that computers would be used for everything, we used to play D&D-type games much as you are describing. One individual–call him dungeonmaster, gamemaster, narrator or whatever–would have created a world (or he might use somebody else’s “canned” world). One or more other players would explore the world; they would tell the DM what they wanted to do, and he would describe the responses of the world to their actions. Generally, the DM would work from previously planned ideas, but he could always ad-lib if the players did something unexpected or the game just went in a different direction than he had anticipated. Interposing a computer to transmit and perhaps interpret the communications between the DM and the player(s) doesn’t really change the basic nature. It’s certainly fun – different than conventional IF, but a valid and fun approach in its own right.