A Modified GOMS Cognitive Task Analysis Technique for Creating Computational Models of Adversary Behavior

摘要

As a result of recent research in cognitive science, the GOMS cognitive task analysis technique has evolved to yield computational models of behavior for many human systems interaction tasks. The types of tasks to which the technique has typically been applied are hierarchical, serial and rather static. The technique produces sets of production rules executable by a production system. Execution of production system models developed as a result of a GOMS task analysis, therefore, follow a rigid flow of control through a defined hierarchy of subgoals. The behavior of such models does not allow for a shift in attention or transfer of control to other subgoals out of the hierarchy during execution. Such models work well for tasks which are rigidly serial in nature and static, in the sense that any change in the task environment is the result of responses made by a single agent. Dynamic task environments, on the other hand, are highly interactive in that the task environment changes not only in response to a single agent but also as a result of the responses of multiple agents and or dynamic physical forces. Such dynamic environments are characteristic of tactical situations and the performance of tactical tasks. Consequently, for modeling adversary behavior for execution in tactical simulation scenarios, the current GOMS approach can not be applied. This paper presents a cognitive task analysis technique that capitalizes upon the systematicity and naturalness of conducting a GOMS task analysis yet introduces modifications to this analysis. These modifications accommodate its application to the analysis and modeling of dynamic tasks with many interacting agents. An example of the knowledge structures produced for computer execution as embodied in a production system architecture is also presented.