A detailed performance record and two types of models of tactical knowledge are produced when students conduct tactical planning and problem solving in the context of a tactical planning simulation, the TAO Sandbox. The detailed performance record includes every action taken by a user in an interactive simulation session. The detailed performance record can be utilized to play back the session in the simulation. This session record includes low-level analyses of session events that are produced by the Sandbox. Our team has developed methodologies for maintaining a model of user knowledge based on these observable performance events.

At the same time that it is recorded, each announcement of a performance event is reported to a collaborating software component, called the CRESST Assessment Application (CAA), which contains a detailed model of user knowledge at a higher level of representation. When a performance announcement is made to the CAA, it informs a Bayesian network that represents the student's exhibited competencies. The CAA, in turn, reports the new network values to a third software component, the CAA Performance Monitor, along with a brief textual message that it received from the Sandbox about the most recent measured performance. The performance 213monitor maintains student records that include complete histories of CAA estimations, together with the Sandbox-generated explanations. The performance monitor can present graphs of changing student models, with support for userselected data filters to support the presentation of charts for different subsets of the Bayes net nodes.