Open Problem: Reusable Gameplay Trace Samplers

Abstract

We identify an open problem in game design assistance and automation: the development of reusable gameplay trace samplers. Inside many sophisticated content generators and design tools is a component that samples interesting and plausible sequences of player actions. Details and summary properties of these samples are used to assess generated content and to inform designers. As the development of this component is technically involved (sometimes comparable to making a second implementation of a game's mechanics), design tools often either make use of entirely custom, game-specific samplers or make due without the ability to sample interesting traces at all. This severely limits the population who could benefit from automation to those who are motivated to develop it for themselves. We propose the development of reusable samplers to ease the development of future design automation tools. This paper reviews several systems that demonstrate the availability of technology required by these samplers and the range of applications they may serve. It also sketches how future samplers might be architected. This proposal identifies one way for technical research to make progress on design automation challenges without making problematic assumptions about the nature of player behavior or designer intent. Filling in this missing infrastructure, we claim, will make the use of artificial intelligence in the design process more accessible and thus accelerate game design projects.