Queen Mary University of London

On the Evaluation of Procedural Level Generation Systems

The evaluation of procedural content generation (PCG) systems currently suffers from a lack of standardized methodology, which hinders the ability to identify genuine advancements in the field.

• Academic research on procedural content generation (PCG) is currently fragmented, as evidenced by an analysis of 137 publications that reveals a lack of standardized evaluation methodologies.
• Researchers frequently prioritize internal metrics over comparative analysis, often to satisfy publication expectations rather than to build a cumulative body of knowledge.
• The field is hindered by a reliance on redundant, custom-built domains and unvalidated fitness heuristics, which prevents effective cross-study comparisons and limits practical industry application.

Hunt Takes Hare: Theming Games Through Game-Word Vector Translation

Automated game design benefits from a novel methodology that bridges the gap between abstract ludic mechanics and semantic thematic content.

• A new methodology enables procedural re-theming of games by mapping game-log embeddings to semantic word vector spaces, decoupling mechanical rules from surface-level narratives.
• This approach preserves the structural and strategic integrity of original game mechanics while allowing for the automated translation of concepts into entirely new contexts.
• Integrating gameplay-specific embeddings with linguistic word vectors produces more contextually relevant thematic outcomes than using word vectors in isolation.

Permalife of the Archive: Archaeogaming as Queergaming

Archaeogaming and queer game studies converge as critical frameworks for analyzing digital environments, emphasizing the ethics of representation, labor politics, and the subversion of hegemonic design.

• The synthesis of archaeogaming and queer game studies shifts the focus from goal-oriented play to games as complex, subjective systems of power and pleasure.
• Players actively resist standardized design by creating personal, incomplete, and messy archives that prioritize lived experience over objective, totalizing histories.
• The conceptual framework of 'permalife' replaces the traditional, masculinist model of 'permadeath,' emphasizing the reparative potential of digital environments.

The Right Variety: Improving Expressive Range Analysis with Metric Selection Methods

Expressive Range Analysis serves as a critical tool for evaluating procedural content generation, yet the field has historically lacked a standardized, objective methodology for selecting the metrics used to visualize generative spaces.

• A new quantitative framework replaces intuition-based metric selection in Expressive Range Analysis with three criteria: Fitness Independence, Mutual Independence, and Minimal Alternative Metric Independence.
• The study validated this methodology by analyzing 9,015 Mario levels across 18 distinct structural and agent-based metrics.
• Combining structural metrics with agent-evaluated metrics produces the most robust visualizations because these categories exhibit lower mutual correlation.

LingoTowns: A Virtual World for Natural Language Annotation and Language Learning

LingoTowns is a web-based, procedurally generated virtual world platform designed to integrate multiple Games with a Purpose (GWAPs) for natural language processing (NLP) annotation and language learning.