Experimental Validation of the Attention Switching Component of the COGNET Framework

Abstract

COGNET has recently been proposed as a general cognitive model of human-computer interaction in real-time, multi-tasking environments. A main feature is a component that models human operator attention shifts as a product of competing task demands and a dynamic external environment. An experimental validation of that attention switching component was undertaken, using a previously reported COGNET model. Separate data were collected from subjects used to build the original COGNET model, and new (but equally expert) subjects. The model was found to predict 90% of all observed task instances (p < .01) for the original subjects, and a surprising 94% for new subjects. For predicted tasks, the model prediction was also found to lead actual task initiation by an average of 3.2 min. for original subjects and 2.2 min. for new subjects (over an average problem duration of 90 min.).