Is this nuclear material secure? Examining trust in voice user interfaces for international nuclear safeguards seal examination

Abstract

Voice user interfaces (VUIs) as decision aids are becoming increasingly popular in both everyday interactions (e.g., mobile assistants on personal phones) and high-risk, high-consequence security settings such as international nuclear safeguards. It is important that users have appropriately calibrated trust in these VUIs. Here, we bridge the domains of international nuclear safeguards, trust in technology, and VUI guidelines by examining human performance and trust in a VUI digital assistant for a safeguards seal examination task. This study serves as the foundation for future work investigating the impact of factors such as explainability, provenance, confidence, and granularity information on user trust in VUIs. This research will help establish best practice guidelines for VUIs within the context of international nuclear safeguards, which may also be applied to other national security VUI applications.