Many digital games—particularly role-playing games—engage players through interleaved story and open-ended gameplay. Players often have different preferences over the content of the game's story and the open-ended adventuring available. We developed GameForge to explore artificial intelligence techniques for procedurally generating, rendering, and making playable novel role-playing games based on a priori unknown story structures.

Generating games requires a balance among meeting player preferences, enabling designer control, and conforming to algorithmic constraints. GameForge allows players to specify which elements of a game story should be included or excluded, employing a plan adaptation approach to modifying the game story. Players also specify their desired level of open-ended exploration (as opposed to direct pursuit of story), influencing a search-based optimization procedure for generating a world map. World generation ensures plot points have locations in the world while adding additional branching and side rewards proportionate to player desires to explore.

Designers constrain the generation process through specifying a model of the game world. This model controls how different game environments connect, the rate of encountering random opponents in the world, and the likelihood of finding treasure in the world. Designers also provide world characters and initial plots for GameForge to use.

GameForge takes the provided information from designer and players to generate a full role-playing game in a pipeline. First, players specify preferences over plot elements in a game story and GameForge adjusts the story to include (or exclude) those elements while maintaining a coherent story. Next, player preferences for exploration and designer world constraints are used to generate a world map that embeds story plot points. Finally, non-player character logic is automatically scripted to be appropriate to the provided story. Together, our approach demonstrates the viability of automatically generating integrated game stories and worlds using artificial intelligence techniques. Developing these methods enables the design of games where players are empowered to adjust content to suit their interests. GameForge was nominated for most innovative project in the AAAI 2011 video competition.

Relevant Publications