Evolved Domination: Exploring generative AI in a turn-based video game context

Abstract

This master’s thesis explores the integration of multiple generative AI models into a turn-based strategy video game to function as a cohesive unit. Generative AI is an interesting topic in the field of video game development since it allows for the automatic creation of content. This can, for instance, increase replayability, shorten development time and enable new game mechanics that were previously impractical. Developments of these benefits have, however, been limited due to the availability of high-quality AI models that minimize hallucinations and inconsistencies when content is generated. To examine this, the strategy video game Evolved Domination was constructed through a feature-focused Scrum workflow. It utilizes 12 AI model instances to generate game data, text, images, music and text-to-speech that form a cohesive experience. The findings are presented through a SWOT analysis conducted during the project, highlighting 13 strengths, 10 weaknesses, 11 opportunities and 10 threats of integrating generative AI in a turn-based context. These findings aim to help researchers and developers understand the challenges and potential benefits of generative AI in games, aiding future research.