Quantitative assessment of language and cultural diversity as human reliability factors in maritime safety

• Quantifies risk pathways of language and cultural diversity in maritime accidents. • Uses Bayesian Network modeling with 550 real-world maritime accident cases. • Identifies key roles of crew diversity, language proficiency, and SMCP compliance. • Sensitivity analysis reveals leverage points for targeted safety interventions. • Provides actionable indicators for policy, training, and crew management. Reliable shipboard operations depend on effective communication and situational awareness among multinational crews. Language barriers and cultural diversity, if unmanaged, can degrade operational reliability and compromise safety-critical decision-making. Yet quantitative evidence on how language and cultural diversity shape accident risk remains limited. This study applies a Bayesian Network (BN) framework to assess the reliability and safety implications of language and cultural diversity using 550 maritime accident investigations. Natural Language Processing (NLP) extracted indicators, including crew nationality composition, language diversity (Shannon Index), cultural diversity (Hofstede Index), and adherence to Standard Maritime Communication Phrases (SMCP), which were integrated into the BN model. Parameters were estimated using the Expectation Maximization (EM) algorithm, allowing the model to quantify pathways to communication errors, conflict, and accident severity. Results show that high language diversity and low proficiency substantially increase communication failures: miscommunication and misinterpretation together account for 68% of observed errors (43% and 25%), with high language diversity the most frequent state (∼48%). Verbal exchanges remain the most failure-prone mode (47%), and container ships account for the largest share of communication-related accidents (42%), with collision and grounding comprising 41% and 24% of incidents. At the consequence level, major repairs are the modal outcome (∼40%), but scenario analysis shows that under highly adverse diversity and proficiency configurations, the combined probability of major repair and total loss can exceed 60%. Strength-of-influence results identify vessel size, nationality, and language diversity as structural drivers (0.54), while language training (0.26), SMCP use (0.24), and cultural familiarization (0.27) are the most effective levers to improve proficiency and reduce conflicts. Scenario analysis further demonstrates that aligning these levers can cut total loss probability on large container ships from about 23% under stressed human-factor conditions to 3% under best-practice communication settings, while shifting incident severity toward minor outcomes and containing casualties. These results provide a quantitative basis for setting language-training thresholds, enforcing SMCP, and designing crew-mix and cultural familiarization policies in maritime safety management.