Operator specific modeling of identification judgments in a complex dynamic task

In this paper, we consider the problem of training operators to make identification judgments from multidimensional, possibly probabilistic, information sources. We used models of the environment and the operator to measure the causes of lack of fit of the operator to the environment. Linear regression was used to model the relationship between cues and identifications. Cues in the environment were regressed on the correct identification to obtain the environmental model. These same cues were regressed on the operators' identity judgments to obtain the operator models. Individual environmental models were necessary in order to properly represent the environment when operator decisions and timing shaped the environment dynamically. High agreement between environmental and operator models were linked to top performance on the task. Implications for training are discussed.