Inferring Design Constraints from Game Ruleset Analysis

Abstract

Designing game rulesets is an important part of automated game design, and often serves as a foundation for all other parts of the game, from levels to visuals. Popular ways of understanding game rulesets include using AI agents to play the game, which can be unreliable and computationally expensive, or restricting the design space to a set of known good game concepts, which can limit innovation and creativity. In this paper we detail how ANGELINA, an automated game designer, uses an abductive analysis of game rulesets to rapidly cull its design space. We show how abduction can be used to provide an understanding of possible paths through a ruleset, reduce unplayable or undesirable rulesets without testing, and can also help discover dynamic heuristics for a game that can guide subsequent tasks like level design.