I want to suggest that there is also an 'Uncanny Valley' of sorts in world-building, that when creating imaginary worlds which feel real to us there is a point where something is uncomfortably almost-but-not-quite real.
(from The uncanny valley of world building, Kit O'Connell)
Much debate follows in the comments, including whether Kit was even understanding his exemplar Morrowind anecdote correctly. Nonethless, a useful idea to apply.
I know, it's on BoingBoing, which means you've already seen it. Nonetheless:
The distributed fiction of I Love Bees was designed as a kind of investigative playground, in which players could collect, assemble and interpret thousands of different story pieces related to the Halo universe. By reconstructing and making sense of the fragmented fiction, the fans would collaboratively author a narrative bridge between the first Halo videogame and its sequel. As the project’s lead writer Sean Stewart explains: "Instead of telling a story, we would present the evidence of that story, and let the players tell it to themselves."
(from Why I Love Bees, Jane McGonigal)
The kind of game design that the creators were exploring will be instantly familiar to fans of the MIT Mystery Hunt:
I would argue that the primary puzzle of I Love Bees embodied a meaningful ambiguity. That is, the data set lacked the clarity of formal interactive instructions, yet maintained a distinctively sensical nature. That is, the choice and ordering of the coordinates did not seem nonsensical. Instead, its arrangement was structured and seemingly intentional enough that it promised to mean something, if only approached in the right way. This meaning was implied through the specificity, volume and overtly designed presentation of the data.
But a Mystery Hunt is largely fixed in form at launch time; it has to be, to allow many teams to compete on a fair basis. The designers may have to fix puzzles on the fly, and perhaps delete some, but they won't usually invent new ones. Certainly not based on a particular team's theory.
I Love Bees, in contrast, was gleefully extended as the (single, universal) team of players made progress. The article goes on to describe how the collective intelligence went way beyond what the creators expected. By the end, the creators were flinging together puzzles that required tremendous feats of player cooperation and networking. The players wound up making dictionaries of game information that were more complete and consistent than what the creators had built. And then the creators started relying on these dictionaries to design later puzzles, and mining ideas from incorrect solving theories...