AI Middleware for Human Social-Cultural Behavior Training Games

For effective training in higher-level decision-making and social-interaction skills, there is a need for computer game technology to provide more realistically human-like non-player characters. As a cost-effective path for integrating greater social-cultural realism into game simulations, we propose a set of AI (Artificial Intelligence) Middleware layers that allow easy extensibility and configuration at a variety of levels of abstraction in game designs. We plan to determine an appropriate game concept and engine, create AI Middleware that is particularly geared toward social-cultural human behavior, and develop an example game instantiation of the design. The implemented game engine will have a relatively open architecture, and plug-in application programmer interfaces (APIs) that provide control of non-player characters (NPCs) in standard ways as well as in ways specific for the modeling of social-cultural human behavior. The system will support alternative human behavior architectures, and provide scripting and behavior-modeling interfaces to allow rapid development of simple and/or well-understood behavior models, but also provide an appropriate migration path for developers (and possibly end users) to improve the models in the areas where more sophistication is desirable.