Computational Models for Near-Real-Time Performance Predictions Based on Physiological Measures of Workload

Abstract Current human sensing and assessment capabilities are limited by an inability to account for the multitude of variables that regulate the human performance state. Monitoring behavior alone is not adequate for prediction of future performance on a given task, and no single physiological measurement can provide a complete assessment that influences performance. This chapter investigates a set of physiological measurements in near-real-time computational models to minimize the number of signals and create highly predictive performance models. We developed multiple predictive computational models to assess when physiology markers that coincide with workload levels are reaching a point where performance decreases or increases may occur in the near future. We found increased accuracy, up to 80%, when the models are adapted to a task, but fitting the model to an individual can cause decrements in its predictive power: when models were created based on a single individual, the computational model's predictive power dropped the accuracy to 70%. We examine the balance between increasing performance by using partially individualized models with the time and effort taken to train a new model. We examined the effects of training models on each session independently and in combination to predict performance on the last session across all participants. We compared these results to individualized models trained on all prior sessions and tested on the final session for each of 35 participants. Finally we assessed each recorded signal’s predictive power through sensitivity analysis within the models.