Every AI assistance aboard a ship creates a new requirement in the officer who uses it: the ability to know when it's right, when it's wrong, and when it's confidently wrong — the literacy of working with a tool that fails fluently. The industry is deploying the systems faster than it's teaching the literacy.
AI literacy for maritime officers isn't optional future skills — it's the supervisory competency the autonomous ship depends on. An officer who can't read the AI's reliability is an officer who can't safely supervise it.
Key findings
- AI literacy is the supervisory competency — knowing what the model can do, where it fails, and when to distrust it. Established
- The core skill is calibrated trust — reliance proportional to the system's actual reliability. Established
- The literacy must cover the failure modes — hallucination, out-of-distribution behaviour, confident-wrong outputs. Established
- It's a distinct competency from seamanship — the AI-era officer needs both, not either. Established
- The syllabus doesn't exist — maritime AI literacy isn't yet an STCW competency, which is the gap. Maritime-GPT analysis
What AI literacy actually is
AI literacy isn't coding — it's judgement about a tool that fails fluently. The literate officer knows the model's failure modes (hallucination, confident error, out-of-distribution behaviour), can read its confidence cues, understands what it was trained on, and knows which questions it's unreliable on. The officer who lacks this literacy treats the AI's output with the same trust they give a colleague's — which is the over-reliance failure the safety case exists to prevent. Established
The syllabus gap
STCW's competencies were written for a human-operated ship — and the AI-supervision skills the new systems demand aren't in the syllabus. The officer supervising an AI navigator needs: what the system is doing, when its confidence is meaningful, how to probe its reliability, and when to take manual control. That's a training curriculum that doesn't exist — the maritime version of the AI-literacy gap that's appearing across every industry deploying these systems. Maritime-GPT analysis
The failure mode it prevents
The literate officer is the human-supervision layer that makes the AI safe — the one who catches the confident wrong answer because they know the system produces them. The illiterate officer is the automation-bias casualty: deferring to the system's fluency, trusting the output's confidence, and missing the error the AI was never going to flag itself. AI literacy is the trust-calibration made trainable — the skill of relying on the machine exactly as much as it deserves. Established
What the competency includes
- <strong>Failure-mode awareness.</strong> What the model does wrong — hallucination, edge cases, confident error. Established
- <strong>Confidence reading.</strong> What the system's uncertainty display means and when to believe it. Research demonstration
- <strong>Training-data context.</strong> What the model was trained on and what it therefore can't know. Established
- <strong>Manual takeover.</strong> How to resume control when the AI reaches its limits. Established
- <strong>Verification habit.</strong> The discipline of checking the AI's output against independent sources. Established
What we know
- AI literacy is the supervisory competency — the officer's ability to calibrate trust.
- The competency isn't currently in STCW — the syllabus gap is real.
- Literacy is distinct from seamanship — the AI-era officer needs both.
What we don't
- Whether maritime AI literacy training measurably improves supervision outcomes.
- The certification pathway for AI-literacy competencies.
- How literacy requirements scale with autonomy degree.
Being tested
- AI-literacy curricula for maritime academies.
- Trust-calibration simulator training.
- Competency frameworks for AI supervision.
Maritime-GPT analysis
AI literacy is the unsolved prerequisite of maritime autonomy — the systems are deploying faster than the workforce that must supervise them is being trained. The literate officer isn't the one who can use the AI; they're the one who knows when it's lying. Until that competency is taught, the industry's AI deployment is running ahead of the human capacity to oversee it — which is the gap the incident reports will eventually fill in. Maritime-GPT analysis
This is a research feature, not operational guidance. Simulation results are reported as simulation; demonstrations are not deployments; company claims are attributed as such. Verify safety-critical decisions against official nautical publications and flag-state requirements.
Sources & evidence
The automation-bias foundation literacy addresses.