Modeling Culturally Accurate Behavior via Multiplayer Game Environments

Military simulations involving non-player characters (NPCs) from non-US cultures are becoming increasingly important for training and mission rehearsal. However, creating culturally-realistic behaviors for these characters remains a technical challenge because there is not a clear standard or process for creating behaviors that are consistent with a given culture. This Phase I effort will design a solution to the problem via a set of tools, called CulturePad, that allow cultural behaviors to be empirically captured and abstracted in such a way that the resulting abstraction can be used to efficiently create accurate models of the cultural behaviors of non-US NPCs. An important objective of the Phase I research is analyzing the potential obstacles to capturing and simulating culturally appropriate behavior, and developing strategies to mitigate these in the CulturePad design. The CulturePad concept is based on a generative narrative representation which has been shown to hold validity as a way of capturing cultural knowledge in multiple cultural settings, and has also been shown to be a feasible way of generating behaviors for computer-generated synthetic agents. This will be accomplished, in part through empirical studies comparing how culture-specific scenarios would be played out in a multiplayer game versus a live ‘role-play’ simulation.