It's not difficult to imagine games where the number of possible unique inputs is far beyond the number of represented rule-creating restrictions within the judgment system. This is the problem with natural language processing - and the reason we do not yet have automated game mastering for roleplaying games.
However, there may be an easy point of entry, here. Easier first steps can be made by creating an artificial umpire who can weigh competing emergent outcomes and make a rule out of the one which would best suit the game.
Compelling 4a games are perhaps the hardest to imagine. Having the computer create a new game for us, even as we sit down to play - this is close in difficulty to the different dream of interactive stories.
As above, however, there may be some easier paths and entry points for approaching this goal. Having a game build a variety of simple game types and respond to what the player prefers seems theoretically possible without first devising a fully human level of AI. We do not have to create the world as a first step. A simple puzzle game will do just fine. We also have the advantage that human players want to help and provide feedback to the system to make it more enjoyable.
If we could realize such a system, and its libraries of rules were well-annotated and significantly generative, would an artificial artist emerge? Perhaps, but the creator of this artificial artist would be a game designer.
The Art of Rules
The possibility of such a system links the knowledge of long-dead designers to our present... and to an unexplored future.
There are plenty of opportunities for all kinds of art, from a massive and grand composition involving millions of players each playing their role, to the artful execution of a single, solitary game as it is created on the fly by an artificial game designer. As we approach this future, the fundamental skill necessary for creating and selecting the right rules grows in importance.
Thankfully, we do not have to worry about such grand speculations to practice the art of rule creation and selection; after all, game designers have been doing it for thousands of years. We merely have to pay attention.
There are many parts of a working computer game that dictate and translate player responses. World geometry, physics, music, character design; it's tempting to regard the game's actual rules as less important, modifying them in support of simulation systems and other game elements. Instead, studying the rules in advance and crafting a message from them represents one of the clearest and easiest creative opportunities for game designer. It is within these rules that the players will inhabit and practice their own art.
Author's note: My thanks to Robin Hunicke for kindly reworking the piece for form. I am also indebted to Charles London for corrections and edits. Finally, Ray Mazza, Matt Goss and Hunter Howe for their insights.
Rod Humble is Vice President and Head of the Sims Studio at Electronic Arts. He has been in the games industry for 15 years as a designer, executive producer and head of studio.