In our daily lives, the artwork we experience is increasingly the output of an algorithm executed within a computer; such algorithms are known as procedural generators. From the humble screensaver on the Macintosh with its swishes and whirls, to large simulated crowds in movies, our experience is enhanced by the power of procedurally generated content. At the same time, specialized computer applications such as Photoshop, Illustrator, and Final Cut Pro have made it much easier for human designers to manually create top-quality content, thereby unleashing the creative potential of millions. Though the cost of tools has come down, the cost of content creation remains high, limiting the volume of content each person can create. The goal of this project is to combine the speed of procedural generators with the expressive potential of manual content generation tools. We seek to create tools where the human designer and a procedural generator work together in a collaborative way, each building on the work of the other. We call such human/computer creative systems ?mixed-initiative,? since at any point, either the human or the computer can take the next step in contributing to the final creative content. With such mixed-initiative content creation systems, the potential exists for a rapid amplification of human creative potential, with the result being increased vibrancy and intricacy across a wide range of creative works.
This project will explore the design space of mixed-initiative content creation systems by constructing and evaluating one such system, called Springboard. Our target area for content creation is levels for computer games, specifically the genre of 2D platformers (such as Super Mario World). Computer games are a representative domain for creative content authoring, since they involve visual arts, spatial layout of game elements, animation, and game mechanics (movement, rewards, win conditions, etc.). Level design hence involves elements found in many existing content creation tools. Building and evaluating Springboard will yield an enhanced understanding of the technical architecture required to create such systems and the kinds of knowledge representations needed to permit flexible interaction of human and computer created content. Also to be explored is the value of creation, storage, and access to variants of the final creative product, thereby supporting rapid exploration of alternate designs. Studies of human designers using the system will provide insights into how mixed-initiative tools influence the human creative process. We anticipate procedurally generated levels will help designers generate more varied ideas during brainstorming activities, will permit designers to explore more design alternatives, and in general lead designers to create longer and more complex final level designs than when using traditional, manual-effort level design tools.
