ABSTRACT
Artificial intelligence (AI) is expected to play a crucial role in enabling ships to operate without direct human input. Due to the safety-critical nature of maritime navigation, AI-enabled systems must exhibit high reliability and robustness across a broad range of scenarios and situations. However, considering the limitations of these systems in handling new and complex situations, careful design, development, implementation, management and operation are necessary when deploying them in real-world environments. As a result, proposed autonomous ship concepts generally involve human operators to monitor, supervise and intervene to ensure the required performance and safety standards are maintained.
Exerting meaningful human oversight of autonomous ships means ensuring system design features are supportive of the cognitive processes needed to perform this task. This chapter discusses three themes from the scientific literature that are central to supervisory performance and addresses how developments within the application of AI in ship autonomy may affect them. These include operators’ mental models of the subsystems of autonomous ships, situation awareness of current and future ship behaviour, trust in the ships’ performance and the operators’ ability to allocate sufficient attention to incoming information.
Finally, this chapter concludes by discussing design strategies for supporting human oversight of autonomous and AI-enabled systems from remote operating locations. Central to this end is applying human-centred design processes whereby goals, decisions, tasks and information needs are systematically addressed in a multi-disciplinary context. Successful implementation of human-centred design strategies will ultimately leverage the team dynamics underpinning human operator and AI collaboration, which in turn hold the potential of their sum being larger than their respective parts.
