Towards an Adaptive Assistance System for Monitoring Tasks: Assessing Mental Workload using Eye-Tracking and Performance Measures

With the introduction of more and more autonomous machines into the work environment, the role of a worker changes from the sole executor of a task to the observer and supervisor of a system that carries out tasks on her behalf. Often, the transparency and predictability of these systems decrease, making it difficult to comprehend underlying processes for the worker. Moreover, monitoring tasks can impose different levels of workload on the human operator leading to an increased risk of making serious errors. The present research aims at developing an adaptive assistance system for these types of tasks that is able to monitor a worker's current level of mental workload and provides support without reducing the worker's autonomy and sense of responsibility. We report results of an experiment using a monitoring task incorporating repeated event sequences to simulate underlying workings of a complex system. Results show that performance in connection with eye-tracking measures are suitable indicators of the level of mental workload and that making the worker aware of underlying structures may reduce workload. Further steps towards an adaptive assistance system for monitoring tasks are discussed.