Human–AI Interface Design for Trust Calibration and Cognitive Workload Management in High Stakes Decision Making Contexts: A Scoping Review

Abstract

High-stakes decision environments continue to face cognitive demands, accountability pressures, and uncertainty, even as artificial intelligence is increasingly embedded in decision support across critical infrastructure, safety-sensitive domains, healthcare, and aviation. This scoping review examines how the design of the human–AI interface shapes cognitive workload, safe reliance, and trust calibration in such settings. A PCC-framed question and a PRISMA-ScR-guided process were used to identify studies published between 2015 and 2025 in Scopus, PubMed/MEDLINE, Web of Science, ScienceDirect, IEEE Xplore, and the ACM Digital Library. These were screened and charted using an extraction template. Seventeen studies were chosen, covering clinical decision support, sepsis management, medical imaging, power-grid congestion management, telehealth diagnosis, air traffic control, medication verification, and maintenance. In most situations, interfaces that combine interactive verification, actionable uncertainty communication, selective transparency, and support for intermediate reasoning were more effective than static explanation designs; however, deployment remains constrained by methodological heterogeneity, limited real-world integration, small samples, limited real-world integration and inconsistent measures. This review proposes a thematic structure linking deliberative support, oversight-preserving design, and calibrated transparency, and offers a roadmap for embedding trustworthy human–AI interfaces in safety-critical decision support systems.