System decision framework for augmenting human performance using real-time workload classifiers

Abstract

The high volume of information available to human operators and increasing scale of work can become unmanageable due to the complexity found in a variety of domains. The need for precise, continuous assessment of human operator performance and state is important to identify when, and how, interventions should be delivered. One challenge that requires attention is the need for intelligent model-driven systems that identify specifically when some form of augmentation is needed while work is performed. Our current research and development efforts seek to fill this need by following the Sense-Assess-Augment (S-A-A) framework. We utilize the Performance Measurement Engine (PM Engine™) and the Functional State Estimation Engine (FuSE2) to derive second-by-second measurements of performance and human operator state to identify the specific points in time where performance decrements occur due to high workload. These human state patterns can be computationally modeled via the Performance Augmentation Cueing Engine in Real-time (PACER) to provide the decision logic necessary to predict when performance decrements are likely to occur. In this paper, we describe the methods used to collect our initial data set and explore the complex relationships between cognitive workload and primary task performance.