I’ve never really written anything about ANGELINA, Mike Cook’s game generating AI and that’s because I’m not sure what I think about it. My folded-arms-and-raised-eyebrow brain finds the goal of the procedurally-generated game as plausible as the procedurally-generated novel. My thoughtful-and-cautious brain sees it more as an ongoing academic project where the journey is more important than ANGELINA itself, baby steps into uncharted territories.
Lots of great work is being done in procedural generation and there’s a good reason I reserved a special section for Cook’s “The Saturday Papers” when I ran Marginalia. (The Saturday Papers seems to be on indefinite hiatus right now.) I guess deep down I was concerned that ANGELINA, for all its creator’s ambition, would end up as the machine-mother for narrow experiences that would merely appear to prove the point of Jesper Juul’s old paper that rules matter most and the rest is just skin deep.
But Cook together with Gillian Smith, wrote a paper titled “Formalizing Non-Formalism: Breaking the Rules of Automated Game Design” which proposes a game-generating AI which eschews rules and win conditions in favour of experience. I know what you’re thinking, that sounds an AI which makes secret boxes. You would be more than right. Electron Dance is actually cited in the paper! Cook and Smith deploy the term secret box instead of walking simulator.
Here’s the abstract:
Automated game design (AGD) is an exciting new frontier for generative software and games research, one which intersects many areas of AI as well as cutting across the many creative domains involved in developing a game. However, there is a trend throughout existing automated game design work to concentrate primarily on the rules that underpin a game–objectives, obstacles, and the notion of challenge. This paper examines this trend in automated game design, and argues that a broader understanding of games is needed. We examine the history of AGD to date, and consider this work in the context of game design theories and definitions. We discuss the term secret box to describe a class of game that does not fall into the purview of existing AGD approaches, and offer a design sketch of an AGD system we are building, ANGELINA 6, to begin to challenge these ideas.
If you’re interested, a PDF of the paper can be found linked from Smith’s site. Don’t worry, it’s fairly accessible as it does not go into technical detail.
Also, congratulations to Cook and Smith as their paper swiped the Best Paper in the Game Design category at this year’s Foundations of Digital Games conference.
Here’s Cook talking about one of ANGELINA‘s entries to Ludum Dare.
You might also be interested in Mike Cook’s presentation The Lost Art of Dreaming.