Generative Methods

This great workshop paper from 2013 argues that we should stop using the term ‘procedural content generation’. Instead, we should refer to these tools and techniques as ‘generative methods’.

The term “procedural content generation” awkwardly suggests that the field’s
output be defined by its ability to produce game “content”, a term that fails
to capture the breadth of artifacts produced by researchers.

I’ve always felt a slight discomfort with the phrasing of ‘procedural content’ and never fully understood what it was supposed to mean. It seems clear that both ‘procedural’ and ‘content’ are the wrong words to describe this kind of creative work. ‘Procedural’ feels too specific to imperative code and ‘content’ elides the distinction between different kinds of artifacts, many of which could be recursively fed back into the generator to produce further artifacts. So it’s good to find a paper which clearly articulates these concerns and provides a better replacement.

This might be more intuitive to me personally, because I spent hundreds if not thousands of hours in my early 20s engaging with this stuff from the perspective of art theory, design science and music, rather than games.

Thinking about generative methods in a broader sense provides an interesting rebutal to Carmack’s dissmissal that “proceduralism is just a truly crappy form of data compression.” There’s probably an element of truth to that, but it’s really an argument in the context of viewport projection of 3D scenes in relation to an unquestioned aesthetic of realism/simulationism. Stochastic processes are much richer and more interesting than that. One gets a distinct impression that these techniques are still broadly viewed by the major game studios as “an indie thing”—I’ve heard stories about generators being pulled out of projects because of higher level fears over material being published that the producers didn’t have direct control over. But none of this uncertainty and doubt comes close to the ridiculous fear and loathing expressed by essentialist music critics over the the supposedly shallow and soulless results of algorithmic music composition.

Need to think further about a taxonomy/shape classification of various generative methods. In a practical sense, it might be useful to follow what web designers have been doing with atomic design—building up from discrete atoms to molecules, compounds, cells, to eventually form whole organisms. Generators embedded inside generators inside generators.