AI for Game Design

The usage of Artificial Intelligence methods can often seriously improve the efficiency of a design process. Techniques such as genetic programming, constraint solving and machine learning are routinely used in areas ranging from the synthesis of drugs to architectural modelling to the design of electronic circuit boards. Nowhere is the need more pressing for improvements in design efficiency than in the games industry. This is because the vastness of game environments, the intricacy of plot lines and the quantity and detail of characters are all taken for granted as part of modern games. The sheer size of games effectively rules out small/medium companies competing in certain genres because the design time and effort is too great. Even the large games companies are beginning to feel the strain. For instance, the EA Games team developing the latest Harry Potter title comprised more than 100 people. Some areas such as automatic landscape generation are fairly advanced, but, compared to other domains, in general AI techniques are barely utilised in the design of video games. There is huge potential to revolutionise this area of the games industry and in doing so provide significant challenges for AI research.

An important, related, area is interactive storytelling. This concerns the dynamic generation of narrative content based on the goals and plans of the agents in a game environment. The aim is to generate content with immersive qualities similar to film, but which unwinds in unpredictable open-ended ways (very unlike film). As demonstrated by the success of the recent Facade game (described as “the future of video games” by the New York Times), interactive storytelling has massive potential to change the gaming world. Recent research has demonstrated some potential, but currently only a small percentage of simulations produce interesting narrative. Some research issues include: automated filtering techniques to extract only the best material; more intelligent non-player characters which regularly produce interesting narrative; improved theory for situation design (e.g., designing mixes of characters such that good narrative is more likely to emerge).