Exploring the value of randomness in procedural generation

“As generative designers we have to think about the whole generative space, not just a single example from it.”

- Mike Cook explores why building procedural generators can be a tricky task.

Game designer and AI researcher Mike Cook has put together a tutorial on his website that explores some basic building blocks of procedural level generation. 

It’s the first of a multipart series on the topic, and should be of interest to developers curious about different types of procedural generation and finding the balance between the unexpected and the fun.

The tutorial explores the relationship generative space, the set of all things an algorithm can generate, and possibility space, the set of all imaginable arrangements, while providing interactive examples of four variations on a procedural level generation algorithm to illustrate those differences.

“This shows you how tricky it is to build a procedural generator. We want to think about our generators like normal game content, to imagine a player playing a specific level and enjoying it,” says Dr. Cook. “But as generative designers we have to think about the whole generative space, not just a single example from it. How big is the space? How rich with surprises is it? How often does it produce something boring, or something bad?”

The full tutorial, and any future entries in the series, can be found over on Dr. Cook’s website.