The emergence of a new class of complex applications in bio-medical and health-care systems, intelligent transportation, disaster situation management systems and others, has defined new requirements to the methods of control of these systems. Central to those applications is the requirement to understand the meaning of complex situations happening in dynamic environments, and to act based upon those situations so that certain goal situations will be reached. Often actions of situation control face hardly definable goal situations and lack of control optimality. Although the importance of theories such as situation awareness has been well recognized, we are still away from a broadly accepted understanding of the mechanisms of situation control. We argue that augmenting situation control with capabilities exhibited by human cognition provides more effective mechanisms for organizing goal-directed behavior of complex systems. The paper presents conceptual framework of cognitive situation control and discusses details of the main components of the proposed architecture, including situation recognition, negative situation control feedback, and action planning.