This insight explores how explainable AI and human-in-the-loop design can close the trust gap in voyage optimisation, using real-world examples from collaborative maritime projects.
Balanced interaction between humans and AI becomes possible only when users can understand, interrogate, and influence system outputs. One example of a successful human-AI balance is the explainable AI (XAI) project with the Alan Turing Institute, designed to ensure AI makes decisions understandable to captains.
The project allows captains to converse with the AI and request constraints rather than simply receiving static recommendations. In a collaborative study with shipping operational optimisation provider Theyr, the Turing Institute addressed the trust gap in explainable AI for maritime routing.
The study found that although the optimisation software could generate fuel- and time-saving routes, many captains were reluctant to follow the recommendations because they did not understand why the system had chosen a specific route. This lack of transparency reduced confidence and limited adoption, even when the outputs were technically sound.
To address this, the project introduced rational visibility. The system showed why a particular route, speed, or manoeuvre had been recommended, linking decisions to weather conditions, fuel-efficiency curves, or vessel-specific performance data. This allowed users to see the trade-offs embedded in each recommendation.
Another key feature was the human-in-the-loop approach. Rather than replacing the captain’s decision, the system augmented it. Captains were able to query the AI with questions such as, “What if I sail at 12 knots instead of 14?” or “What if I deviate from port X?” The system then displayed comparative outcomes, enabling informed judgement rather than blind acceptance.
Feedback loops were also built into the system. Captains and shore teams could provide feedback on why recommendations were accepted or rejected, allowing the model to improve over time. Contextual explanations were tailored to users, ensuring explanations matched the user’s experience and operational role.
As a result, a voyage optimisation system developed through this approach achieved around 5% fuel savings compared with existing routing solutions and improved Time Charter Equivalent by around 8% in one case. These outcomes demonstrate the tangible benefits of combining AI’s optimisation capability with human understanding and accountability.
To explore how a balanced human-AI approach can strengthen voyage optimisation in safety-critical conditions, read the full Thought Leadership report ‘Co-Pilots of the Sea: Exploring the Human Intelligence Behind Maritime AI’ produced in partnership with StormGeo.
