Source-linked AI summary
Understanding Operator Attitudes Toward AI-Supported Decision Making in Maritime Operations
Doreen Jirak, Armeen Saroukanoff, Dirk van Rooy
TL;DR
Safe integration of MASS and AI-supported decision assistants depends on how maritime professionals perceive and trust these systems. This study surveyed stakeholders using collision-avoidance scenarios, questionnaires, and open-ended feedback, finding generally positive attitudes alongside concerns that support calibrated rather than unconditional reliance.
Problem
Maritime autonomy shares decision-making with human operators, creating a need to evaluate whether AI systems fit maritime responsibilities, practices, and safety-critical operations.
Method
The study surveyed seafarers and maritime stakeholders using two collision-avoidance scenarios, established and adapted questionnaires, and open-ended feedback analysis.
Results
Participants showed generally positive attitudes toward maritime technology and AI-supported assistance, with comparable technology openness across age groups and broadly stable trust across scenarios.
Takeaways & Limitations
Maritime AI should support calibrated reliance through reliable, transparent, context-sensitive explanations and continued involvement of human expertise.
Takeaways & Limitations
The limited final sample and static scenario images restrict generalizability and may have reduced realism or immersion.
Abstract
from arXiv · showhide
Maritime Autonomous Surface Ships (MASS) and AI- supported decision assistants are expected to transform maritime operations, but their safe integration depends on how maritime professionals perceive and trust such systems. This paper presents a survey study on maritime stakeholders' attitudes toward an AI-supported assistant in collision-avoidance scenarios. Participants evaluated technology anxiety, trust in automation, and explanation quality using established and adapted questionnaires, complemented by sentiment and thematic analysis of open-ended responses Results indicate a generally positive disposition toward maritime technology, no clear age-related differences in openness, stable trust across scenarios, and more scenario-sensitive, multidimensional explanation ratings. Open responses showed that participants valued support for decision-making, situation awareness, and confidence-building, while raising concerns about AI reliability, over- reliance and loss of expertise. The findings suggest that maritime AI systems should not focus solely on increasing automation or trust, but on supporting calibrated reliance through transparent, reliable, and operationally meaningful design with domain experts in the loop.
1 Introduction
Maritime autonomy redistributes decision-making between human operators and intelligent systems rather than removing humans. The study examines how maritime stakeholders perceive trust, technology anxiety, explanation quality, and the advantages and disadvantages of an AI-supported assistant.
- 1 Introduction: Increasing autonomy changes operators’ roles toward monitoring, interpreting, supervising, and intervening when necessary.The transition creates tension between reducing workload and requiring operators to understand, trust, supervise, and override autonomous systems.
- 1 Introduction: Safe collaboration requires calibrated rather than unconditional trust, especially in collision avoidance, COLREG interpretation, and uncertain sensor conditions.
- 1 Introduction: Explanations may clarify recommended maneuvers, traffic interpretation, rule compliance, uncertainties, and alternatives, but can increase cognitive load when poorly timed or misaligned with operational context.
- 1 Introduction: The study examines maritime stakeholders’ trust, technology anxiety, explanation quality, and perceived advantages and disadvantages of a maritime assistant.
- 1 Introduction: The evaluation aims to inform human-centered maritime AI by highlighting challenges in measuring trust and explanation needs in a safety-critical domain.
2 Explainable AI in Collision-Avoidance at Sea
The survey used two collision-avoidance tasks and questionnaires to study seafarers’ trust, technology attitudes, explanation perceptions, and expectations of an AI maritime assistant. Responses generally aligned with the intended scenario interpretations.
- 2 Explainable AI in Collision-Avoidance at Sea: The survey targeted seafarers with bridge command experience and used two collision-avoidance tasks to examine human-centered AI needs.
- 2 Explainable AI in Collision-Avoidance at Sea: The research questions addressed technological progress and trust, explainability’s effects on satisfaction and adoption, and perceived usefulness, trustworthiness, concerns, and expectations.
- 2 Explainable AI in Collision-Avoidance at Sea: The online survey was pilot-tested with ten professional seafarers before dissemination to maritime professionals and nautical students through Qualtrics and an alumni network.
- 2.1 Survey Flow and Tasks: Scenario 1 depicted a regular encounter with a clear procedure, whereas Scenario 2 presented ambiguous traffic involving multiple nearby vessels, a fishing boat, and a buoy.
- 2.1 Survey Flow and Tasks: Participants first assessed ECDIS and AIS images independently, then viewed simplified XAI showing key rudder and speed features alongside the agent’s decision.
- 2.2 Survey Collection and Demographics: Of 166 survey accesses, 66 completed responses remained after filtering, with an uneven demographic distribution and approximately half reporting at least 72 months of sea service.
- 2.2 Survey Collection and Demographics: Participants’ observations and decisions generally aligned with the intended scenario interpretations because the scenarios contained no unrealistic or deceptive elements.
3 Results
Participants were generally open toward maritime technology, while trust remained stable across collision-avoidance scenarios and explanation perceptions varied across items. Open-ended responses supported AI assistance but identified reliability, over-reliance, and interface-design concerns.
- Questionnaires Evaluation: Most participants disagreed with technology-anxiety statements, indicating generally positive attitudes toward technology.The predominant ATAS median was 2, although selected items showed greater variance and concern about technology integration.
- Questionnaires Evaluation: Trust in automation did not differ significantly between scenarios, indicating stable overall trust despite increased traffic complexity.The paired-samples t-test found no significant difference between scenarios (p = 0.936 > 0.05).
- Questionnaires Evaluation: Participants showed moderately favorable but cautious perceptions of the assistant, with median 2 for adverse characterizations and median 3 for positive trust statements.The pattern suggests trust was neither strongly negative nor fully consolidated.
- Questionnaires Evaluation: Technology anxiety did not differ significantly across age or sea-experience groups, and trust differences by age or experience were not statistically robust.For trust, age-group tests yielded p = 0.962 and p = 0.062 across the two scenarios; sea-experience tests yielded p = 0.380 and p = 0.216.
- Questionnaires Evaluation: Explanation ratings reflected multidimensional perceptions, with reverse-coded item 1 averaging 4.1 for ease of understanding and only minor scenario differences overall.The adapted explanation scale showed moderate internal consistency, with α = 0.633 for scenario 1 and α = 0.609 for scenario 2.
- Qualitative Analysis and Synthesis of Open Format Responses: Open responses valued AI support but raised concerns about faulty inputs, complex traffic situations, over-reliance, distraction, and alarm burden.Sentiment analysis indicated that negative comments reflected conditions under which AI support could become unsafe rather than general rejection of AI.
4 Discussion
The discussion interprets the survey as evidence that maritime AI acceptance depends on operational fit, not technical capability alone. It supports calibrated trust through reliable, transparent, context-sensitive systems while recognizing limits from the small, uneven sample and static scenarios.
- Discussion: Participants recognized AI assistants as potentially supporting decision-making, situation awareness, and operational efficiency while remaining broadly positive toward new maritime technologies.The findings were based on perspectives from domain experts and experienced maritime personnel.
- Discussion: Maritime AI development should target calibrated trust through reliable, transparent, and context-sensitive behavior and explanations rather than unconditional trust.The discussion connects this requirement to making system capabilities, limitations, and uncertainties understandable to operators.
- Discussion: The analysis found no clear age-related differences in technology openness, suggesting age alone is insufficient as a predictor of acceptance or skepticism.The discussion identifies perceived usefulness, safety-critical responsibility, and familiarity with bridge systems as domain-specific factors that may matter more.
- Discussion: Generalization is constrained by inconsistent responses, substantial attrition, a limited final sample, and potentially low realism from static scenario images.The paper proposes more dynamic scenario videos or lightweight demonstrations and continued co-design with domain experts.
5 Conclusion
Maritime stakeholders are generally open to AI-supported decision assistance but remain attentive to reliability, over-reliance, distraction, and loss of expertise. The study therefore supports calibrated reliance through transparent, reliable, operationally meaningful design that keeps human expertise central.
- Maritime stakeholders are generally open to AI-supported decision assistance while remaining attentive to reliability, over-reliance, distraction, and loss of expertise.
- Future maritime AI systems should support calibrated reliance rather than focus solely on increasing automation or trust.
- Transparent, reliable, operationally meaningful design and continued human expertise should guide the development and evaluation of maritime autonomous systems.