Traditional human performance models have often been criticized for failing to represent and predict goal-oriented behaviors, and for failing to predict measures that are meaningful to other training and equipment simulations. To address this criticism, in 1999 the Air Force Human Research Laboratory began an effort to develop a human performance modeling environment that could interact with other simulations using an HLA-compatible protocol. One element of that environment is a model development tool that enables users to create a detailed simulation of a goal-oriented human agent, operating in a complex environment. In this context, the simulation predicts what the human is likely to do next based on the currently relevant goals, and on the status of other parallel simulations. A practical example is a combat pilot who has a primary mission to conduct reconnaissance of a target area. Therefore, the pilot's original goal is to fly a well-defined path and to use a variety of sensors to collect data. However, if during that flight the pilot identifies an incoming threat (from a parallel radar simulation), the goal will change immediately to "evade and survive." This dictates a change in tasks as the pilot suspends his execution of the pre-planned flight path and begins new tasks to dump chaff and to conduct high-speed maneuvers.

This is an extremely dynamic and demanding modeling challenge, because goal states change based on events in the scenario as well as on occurrences experienced by agents in other linked simulations. For this reason, they cannot be scripted. The problem is also complicated by the interaction between goals, in which a high priority goal can suspend, halt, or restart a lower priority goal. This must be accomplished with as little burden on the user as possible through the automatic exchange of data and the implementation of sophisticated algorithms to mediate competition between active goals.