Towards the integration and evaluation of online workload measures in a cognitive architecture

Abstract

Adapting an automation system to the workload level of a human operator can be beneficial in many situations, e.g. at industrial workplaces or in safety-critical situations like driving and flying an aircraft. However, this requires real-time assessment of workload. We present a model-based approach for online simulation and assessment of cognitive workload, based on analysing the activities of a cognitive architecture during simulation. A driving simulator experiment was used to evaluate the approach. The cognitive workload of participants was manipulated with a variant of the n-back task as secondary task that parametrically varies memory workload. A virtual driver model was created using the cognitive architecture. The model was simulated in the same situations as the human drivers. The activities of the cognitive architecture as indicator of cognitive workload increased with increasing difficulty of the n-back task. To relate model workload to human driver workload, we compared it with neurophysiological workload measures based on functional near-infrared spectroscopy (fNIRS) brain activation as a complementary measure.