State modeling to identify critical situations

A method for modeling complex human-integrated systems is presented, in which 'critical situations,' which are system configurations where a possible but unexpected control action by an agent in the system can result in an accident in less time than agents need to detect and respond, can be identified. The models are developed from the perspective of the system, and can be demonstrated to be complete. It is argued that, in order to ensure a system is safe, these situations must be controlled, but in many cases such an assurance would render the system impractical or even infeasible. Lastly, it is shown how this method can be used to evaluate the effectiveness of a system, in addition to its safety.