Toward a Tactical Pilot's Assistant: A Framework for Integrating Models of Intelligent Behavior

摘要

Historically, cognitive modeling has been an exercise in theory confirmation. Cognitive architectures were advanced as computational instantiations of theories that could be used to model various aspects of cognition and then be put to empirical test by comparing the simulation-based predictions of the model against the actual performance of human subjects. More recently, cognitive architectures have been recognized as potentially valuable tools in the development of software agents by providing intelligent behaviors that can either reproduce or support human performance in complex domains. While the introduction of cognitive architectures to what has been regarded as the exclusive province of artificial intelligence is a welcome turn, the traditional methods of cognitive modeling are hard to ignore. In particular, there is a tendency to apply cognitive architectures as monolithic, one-off solutions. This runs counter to many of the best practices of modern software engineering, which puts a premium on developing modular and reusable solutions. In this paper, we describe the development of a novel software infrastructure that supports interoperability among cognitive architectures. Our motivation is to allow the analyst to choose the right tool for the job, so to speak, by promoting the decomposition of agent behaviors so that cognitive architectures can be applied in a piecemeal manner with the requirements of the behavior driving the choice of architecture. In this way, we stand tradition on its head and compel the analyst to think carefully about the various ways a behavior might be modeled and, by extension, how those models are best encapsulated so that they can interoperate and, potentially, be reused. After describing the infrastructure, we report on an ongoing effort to use it for the development of a complex "tactical pilot's assistant." We are collaborating with researchers at Soar Technology and Eduworks to understand how otherwise distinct agent technologies might be combined. Specifically, we are exploring how a model of agent perceptions can be combined with a diagnostic reasoning module to assist a pilot in generating and verifying expectations about a tactical situation. In addition to overcoming the practical challenges of model integration, this effort entails a richer theoretical challenge of determining a useful division of labor among agent models and the human operator they are designed to support.