Theory and practice of designing generative AI games: an autoethnographic case study

Abstract

Generative AI (GenAI) technologies are opening up new possibilities in game design, potentially bridging the long-standing gap between tabletop role-playing games (TTRPGs) and digital games. While digital games have excelled in areas such as graphics and complex simulations, they have struggled to match the narrative flexibility and open-endedness of TTRPGs, largely due to the constraints of authorial burden and predetermined content. This thesis addresses the challenge of designing games that combine TTRPG-like narrative freedom with digital game mechanics using generative AI. Through an autoethnographic case study, I show both the practice of designing generative games in a case study at a game studio, as well as the development of a new theory and design framework for generative AI games. This framework, grounded in practical experience, introduces the MAS (Mechanics, Agents, Significs) pillars, the 4F (Function, Fiction, Form, Flow) model of outcomes, and the POV (Possibilities, Operation, Virtual world) model. These findings provide valuable insights into the unique considerations of designing generative games, including technical challenges, player experiences, and the integration of GenAI. By offering a detailed lens into the theory and practice of designing generative AI games, this research lays groundwork for this emerging field of game design.