Evaluating the Effect of Different Multitasking Conditions in AI-Supported Attention Management Systems

Abstract

Monitoring and switching between multiple dynamic tasks is prevalent in various operator scenarios like automated driving, robotics, or aviation. Attention Management Systems (AMSs) aim to support users in these activities. Recent works have highlighted the potential of reinforcement learning-based AMSs, but so far, the approach has been evaluated only in a limited set of multitasking scenarios. This paper describes various factors relevant to multitasking and calls for a systematic revaluation of attention management using such criteria. Thereby, we present a first experiment that modifies some of these factors. Our results indicate that attention management, as developed so far, could not benefit users in all situations. Nevertheless, the study reveals other important criteria that have the potential to improve operators’ multitasking performance.