Modeling human operators in constructive simulations through agents

The article describes a novel approach wherein software agents observe human operator actions in an interactive simulation environment, and use genetic algorithm-based learning methods to map human operator actions to the state of the world with which the operator interacts. The state information is primarily about the environment and cockpit presented through the interface in the interactive simulation environment. The context is information processing and problem solving in a complex, dynamic, and hostile environment. The paper describes the overall approach, the components of a computational architecture, the problem scenario modeled, and preliminary results of the application