Failure in Human-AI Interaction

In the maritime industry, automation and big data are accelerating AI adoption; however, the opacity of complex models hinders integration in safety-critical areas like autonomous vessels. This has spurred growing interest in explainable AI (xAI). Yet, the user experience (UX) of current xAI frameworks remains underexplored, highlighting the need for designs that enhance cognitive ergonomics and interpretability (Dieber & Kirrane, 2022). While automation can reduce certain human errors, it can also introduce new ones (Lützhöft & Dekker, 2002), and, for Maritime Autonomous Surface Ships, the evolving risk landscape remains only partially understood. This project investigates how users interact with autonomous systems through studies in real and simulated environments. A PhD student at the brAIn lab will identify key challenges in human-AI interaction and conduct experiments on its cognitive and emotional aspects, using advanced psychophysiological measurements and modelling. The goal is to provide evidence-based insights into how automation and data exploration influence attention, response time, and frustration, and how system design can enhance these factors and reduce errors.