Designing Games around AI
Artificial intelligence techniques are often designed to solve specific problems of existing games such as Chess, Go, or Starcraft. Reversing this paradigm and designing games to highlight specific artificial intelligence techniques has received some attention but still constitutes a promising future direction for both AI and game development. This tutorial will discuss current attempts at designing games around AI, both in the industry and in academia, elaborate on design patterns for AI-centered game design, and brainstorm ways of building games around several popular AI techniques in a way that puts the AI at the forefront.
Antonios Liapis is a Lecturer at the Institute of Digital Games, University of Malta, where he teaches game design and procedural content generation. His research interests revolve around the interaction between human creativity and computational creativity, for which he has developed several AI-assisted game design software such as the Sentient Sketchbook level design tool.
Michael Cook is a Senior Research Fellow at the University of Falmouth Games Academy where he researches computational creativity and procedural generation. He is the founder of the Procedural Generation Jam and the designer of ANGELINA, an automated game-designing AI.
Intrinsic motivation in general game-playing and NPCs
To date, most agent controlling methods, including predefined behaviours or machine learning, focus on optimising towards explicit goals or fitness functions. This poses several challenges to adaptivity and generality. In contrast, intrinsically motivated agents adapt to changes in the game world and produce behaviour meaningful to the agent themselves, thus addressing some of the challenges imposed by sandbox games such as Minecraft or Factorio, or recent developments in PCG. In this tutorial, we will introduce intrinsic motivation from both a Psychology and AI perspective. While we will present several existing formalisms, and example applications to games or game-like environments, this tutorial focuses on the information-theoretical framework of empowerment, and its application in general game playing and the design of companion NPCs.
Christoph Salge is a Research Fellow at the University of Hertfordshire, working on socializing Sensor-Motoric Contingencies in the H2020 socSMCS project. Christoph does research in Artificial Intelligence and Artificial life, where he is using Information Theory to formalise the drives and motivations of agents.
Tobias Mahlmann is a Postdoctoral Researcher at the University of Lund. Tobias’ work covers various aspects of computational intelligence applied to games, including PCG, player metrics mining, or agent behaviour methods.
Christian Guckelsberger is a PhD student in the Computational Creativity Group at Goldsmiths, University of London. He employs formal models of intrinsic motivation to design agents which act creatively in their own right. His recent work uses intrinsic motivation to drive general companion NPCs and procedural content generation.