The principal way to improve the human-machine system is to work at the level or human contribution. The use or a cognitive model of human behavior for sustaining the design or the study of the interaction with the system under control is increasing. A cognitive simulation model (COSIMO) is described, which simulates the behavior or an operator controlling a complex system during the management of accidents. Particular attention is paid on the theoretical foundations of the model, on its computational implementation and on a number of simulations cases of man-machine system interactions. The approach followed in COSIMO is to build a structure for the various kinds of cognitive functions that are performed by an operator in complex environments, e.g., information seeking, pattern recognition and diagnosing, monitoring, planning and acting on the system. For the implementation of COSIMO a blackboard architecture has been chosen. This architecture offers enough flexibility and modularity for modeling the overall state of the human and of the environment.
