TLDR: This research paper proposes a user-centered approach to integrating Explainable AI (xAI) into maritime decision support systems. It highlights the critical need for transparency and interpretability in AI to build trust among maritime professionals, especially given the industry’s unique challenges and safety-critical nature. The authors present the design of an empirical survey to assess seafarers’ perceptions of trust, usability, and explainability in AI-assisted navigation, aiming to guide the development of user-centric xAI systems tailored for the maritime domain.
The maritime industry is on the cusp of a significant transformation, with autonomous technologies increasingly shaping operations at sea. As Artificial Intelligence (AI) systems become more prevalent in critical decision-making, understanding the ‘why’ behind an AI’s recommendation is becoming as vital as the decision itself. This is particularly true in the complex and dynamic environment of maritime operations, where trust in AI hinges not just on performance, but also on transparency and interpretability.
A recent research paper, “From Sea to System: Exploring User-Centered Explainable AI for Maritime Decision Support”, by Doreen Jirak, Pieter Maes, Armees Saroukanoff, and Dirk van Rooy, delves into the crucial role of Explainable AI (xAI) in fostering effective human-machine collaboration in the maritime domain. The authors emphasize that informed oversight and a shared understanding between humans and AI are essential for safety and efficiency.
Unlike the aviation sector, which has a long history of automation, the maritime industry is known for its conservatism and decentralized nature, making the integration of autonomous systems uniquely challenging. Operational decisions at sea are influenced by a multitude of factors, including weather, team dynamics, and long-standing seafaring practices. This context highlights a potential ‘trust gap’ between maritime professionals and AI systems. xAI offers a promising solution by making the rationale behind AI-driven decisions transparent and interpretable, thereby building user confidence and supporting safe collaboration.
The paper introduces the design of an empirical study aimed at evaluating maritime stakeholders’ attitudes toward explainable AI in navigational decision support systems. The core objective is to investigate how explainability features impact user trust, perceived usefulness, and willingness to engage with AI-assisted systems. This involves a survey-based experimental framework that uses scenario-based stimuli, pre and post-questionnaires assessing trust, openness to technology, and user satisfaction, along with interactive elements grounded in realistic radar-based tasks.
The research is guided by three main questions:
Seafarer Disposition and Trust
The first research question explores the seafarer’s general disposition towards technological progress and their inherent propensity to trust new systems. This is assessed through a pre-questionnaire before any AI interaction.
Impact of Explainability Features
The second question investigates how the inclusion of explainability features, such as those from a ‘maritime assistant’ displaying its decision and the most significant factors involved, affects user satisfaction, trust, and ultimately, the willingness to adopt such systems. This is measured through post-questionnaires after scenario presentations.
Also Read:
- Navigating the Complexities of Human-AI Value Alignment: A Comprehensive Review
- ODiSAR: Digital Twins for Proactive Anomaly Detection in Self-Adaptive Robots
Perceptions, Concerns, and Expectations
The third question aims to understand how maritime professionals perceive the usefulness and trustworthiness of the maritime assistant. It also seeks to uncover their concerns and expectations regarding the integration of AI into daily maritime routines and team workflows, gathered through open-ended questions in the post-questionnaire.
The experimental flow of the survey includes a briefing, the main survey with scenarios (e.g., collision avoidance using radar images), the introduction of the ‘maritime assistant’ with xAI explanations, and a debriefing where demographic data is collected. The goal is to inform the development of explainable, trustworthy AI systems specifically tailored for the maritime domain, providing early insights into human-AI interaction challenges and guiding future research.
In conclusion, the authors advocate for moving beyond purely technical considerations when introducing intelligent systems into the maritime domain. They emphasize the need to incorporate human factors such as cognitive resources, affective states, trust, and willingness to adopt technology. By focusing on user-centered xAI, this research aims to raise awareness among seafarers about technological changes and encourage AI developers to build user-centric interfaces that facilitate meaningful AI integration, especially within dynamic maritime teams. The ultimate success of hybrid learning and decision-making systems at sea will depend on their alignment not only with performance metrics but also with the lived experience and expertise of those who operate them.
