Main features of the human supervision of a long response latency process

A research program on blast furnace supervision has led to better understanding of the main features of human supervision of a very slow process (with time lags varying from some minutes to several hours), with very indirect access to process information. The implications as regards modelling of operational knowledge representation (integrating causal, functional, and topographical viewpoints), planning, diagnosis and prognosis activities, and time processing are discussed. The data also shed light on the nature of human expertise and the value of human-like computer support to human supervision.<>