Crimes Against Mimesis is a series of web articles I wrote in 1996 about the representation of reality in text adventure games, or "interactive fiction". It's gotten a certain amount of fame, even cited in Wikipedia, despite my leaving the IF community shortly thereafter without really having done anything other than writing those articles. Ironically, the text adventure I had been preparing and left half-finished was more a collection of puzzles than anything literary. Its first puzzle, "Cube and Key," was actually an adaptation of an RPG puzzle room I'd come up with earlier on. So it's only full circle to bring these observations about puzzles back to the RPG context.
In-game puzzles and problem solving are one big way to answer the call for more player skill involvement in what's come to be called the Old School Renaissance of Dungeons & Dragons. Even running a system neck-deep in character skills, I doubt if there was ever a DM so bold as to create a riddle and then make the players roll dice against their character's Intelligence to see if they solved it. For all the huffing and puffing about dimwit characters "cheating" by solving puzzles with their players' true IQ, this is one area where the players have always refused to abdicate responsibility.
The big difference between programming a computer and running a live game? Many of the unrealistic aspects of puzzles I objected to in Crimes are irrelevant. The DM understands all the command words the player can use, has a mental model of the world that can tell the players how their attempts are working out, and if any area of the mental model is hazy the DM can consult the oracular dice. The items available with which to manipulate the world do not have to be pre-set and scattered around, but can be bought in town, manufactured, or improvised.
One wide-open area of the tabletop RPG, too, is the possibility for player solutions that surprise the DM. It's an even more gutsy move to put a problem in the way of the players that you don't know the solution to yourself. But really, any interactive problem takes a certain combination of faith in the players to solve it, and faith in the robustness of the game to survive the consequences of their possible failure to solve it.
In Crimes one of my main pleas (in part 4 of the essay) was for text adventure puzzles to become more like real-life problems - goals that can be reached by several different means, or by combining several different objects in a way that makes sense. Even if implemented with real-world objects, a lock-and-key puzzle is still a lock and key. You need the broom to frighten away the cat so you can reach in the mousehole and get the flashlight. You can't frighten away the cat with the crowbar - so says the computer program, but a good GM (or for that matter a good programmer) will increase the realism of the world simulation so that you can.
Even with all these possibilities, many old-edition D&D adventure designers didn't (and don't) take full advantage of them. The game has a good system for combat, and improvised actions and environment effects can be taken into account using it. But there's less of a vocabulary, less of a system, less of a precedent for handling non-combat problem solving. We all know a good puzzle or a tricky problem when we see it; but how do you get there in the first place?
I'd like to know your thoughts before I write on...