Melete: Play testing and 3D environments for Mixed Initiative Artificial Intelligence as a method for prototyping video game levels

Recent advancements in artificial intelligence and human-computer interaction have led to tools that help human authors collaborate creatively with computing systems. Many such tools are available for 2D video games, but exploration of these techniques in a 3D environment is necessary. This paper introduces Melete, a mixed-initiative artificial intelligence tool which uses the Wave Function Collapse algorithm and a human-in-the-loop design process to create and test 3D levels and environments for video games. Analysis by a panel of expert game developers in two design challenges—a battle royale game and a real-time strategy game—indicate that Melete enhances designer engagement and facilitates iterative design. However, it also draws attention to the importance of play testing in the design loop, kernel generation and the usability of such tools. The findings suggest that Melete’s workflow helped designers to seed ideas, explore concepts, and refine implementations quickly with creative agency, but also suggest that further work is needed to integrate Mixed-Initiative Artificial Intelligence techniques into design tools effectively.