Visualizing AI-Supported Adaptivity in Command and Control Interfaces

Abstract

The user interface of Command and Control (C2) systems allows the radar operators to interact with the system. It enables them to surveil, assess, and detect anomalies in the air, on land, and at sea, thereby gaining situational awareness. The systems are complex, processing vast amounts of information, which is reflected in the user interface. The large volume of information, in combination with the high workload demands, risks causing cognitive overload. This could potentially lead to degraded task performance and fatigue, which, in high-stakes contexts, can have serious consequences. To address this issue, this thesis, conducted in collaboration with Saab Surveillance, investigates where AI-supported adaptivity can be implemented in C2 systems, and how it should be visualized. The effects of the visualizations are investigated under varying workload to draw conclusions about efficiency and the operator’s mental workload. The project achieved this through an iterative Research through Design approach, where user interviews and research of the systems led to the creation of visual prototypes within three categories of concepts. These were called Filter, Target Prio-List, and View. All of these concepts were evaluated with participants with operational experience. The results of which provided relevant feedback, showing various potential with all concepts, including possible improvements and future research. An Adaptivity & Workload Test was developed and conducted, investigating operators' efficiency and Situational Awareness under various workloads. These were performed on interfaces with and without an adaptive Filter, comparing the results. The results from the test indicated that, with a low workload, performance and Situational Awareness were similar between interfaces with and without a filter; however, participants perceived their control to be higher without the filter. For a high workload, the performance and Situational Awareness were better with the adaptive filter, as well as the participants’ feeling of control. Concluded, the results indicated that an adaptive filter aids the operator under high workload. Target Prio-List and View were also seen to have potential, however, they need to be developed and tested further, similar to Filter, to be properly evaluated.