INRAT-H cheat sheet

Skill-based slips and lapses are execution failures on routine tasks; mistakes are planning failures (rule-based or knowledge-based) where the intended action itself is wrong.

Rasmussen's SRK model: skill-based (automatic, sensorimotor), rule-based (stored if-then procedures), knowledge-based (novel reasoning); error likelihood and type shift across these levels.

Reason's generic error modelling: slips (attention failure), lapses (memory failure), mistakes (rule-based: misapplied good rule, bad rule, or knowledge-based: incomplete/incorrect knowledge).

Violations are deliberate deviations from rules or procedures and are distinct from errors; routine, situational, and exceptional violations have different causal roots.

Performance shaping factors (PSFs) include time pressure, workload, fatigue, training, procedures, HMI quality, supervision, and environment; PSFs multiply or attenuate error probability.

Human error probability (HEP) is the likelihood a specified task step fails; in THERP-style methods HEP is modified by multiplying nominal HEP by PSF/performance-shaping multipliers.

Error recovery depends on detection opportunity: errors with short detection windows and no independent checking have higher consequence probability than errors caught by self-check or cross-check.

Latent conditions (design, maintenance, training, procedures) lie dormant in the system and combine with active failures at the sharp end to breach defences.

Working memory capacity is limited (commonly cited around 4 chunks, span 7±2 for simple items); concurrent tasks competing for the same resource degrade performance.

Wickens' multiple resource model: resources are partly separate by processing stage (perceptual-cognitive vs. response), modality (visual vs. auditory), and code (spatial vs. verbal); tasks sharing resources interfere more.

Signal detection theory: sensitivity d' separates true discrimination ability from response bias (criterion); hit and false-alarm rates determine both.

Vigilance decrement: detection of rare, low-salience signals declines over time on watch, especially under low event rate and monotonous conditions.

Change blindness and inattentional blindness: unattended changes or unexpected objects may go unnoticed even when in the visual field; expectation guides perception.

Recognition is generally more reliable than free recall; checklists and displays that support recognition reduce memory-load errors.

Prospect theory: losses loom larger than equivalent gains (loss aversion), and probability weighting distorts rare-event perception; this affects risk decisions under uncertainty.

Naturalistic decision making (RPD): experts match situations to recognized patterns and evaluate options serially rather than comparing all alternatives; expertise is pattern-based.

Alarm rationalization targets: every alarm should be meaningful, prioritized by consequence and time to respond, and have a defined operator action; flood and nuisance alarms are design defects.

Alarm prioritization commonly uses consequence (safety/environment/production) and available response time; high-priority alarms should be few and actionable.

Compatibility principle: display and control movement should match operator expectation (e.g., up = increase, clockwise = increase) to reduce mapping errors.

Population stereotype and control-display relationships: spatial mapping (location congruence) and movement congruence reduce errors; violations increase slips.

Situational awareness (Endsley) has three levels: perception of elements, comprehension of meaning, and projection of future state; design should support all three.

Colour should not be the sole coding dimension; redundant coding (shape, position, text) supports colour-vision-deficient users and reduces confusion.

Display hierarchy should match task: overview first (plant-level), then detail on demand; excessive data density increases search time and missed signals.

Control room layout should support team communication and supervision: line of sight, shared large displays, and proximity of related controls reduce coordination cost.

Procedure types: normal operating, abnormal/emergency, alarm response, surveillance, and administrative; each has different verification and usage requirements.

Procedure usability criteria include action steps, decision logic, cautions/warnings, place-keeping, and clear entry/exit conditions; ambiguous steps are a common error source.

Task analysis methods: hierarchical task analysis (HTA) decomposes goals into subgoals and operations; cognitive task analysis (CTA) captures decision and knowledge demands.

Link analysis maps frequency and importance of operator interactions with displays and controls; high-traffic links should be short and well supported.

Human factors engineering (HFE) integration should occur early in design; late HFE changes are costly and often limited to training or procedure fixes.

Verification vs. validation: verification confirms the design meets specified requirements; validation confirms it supports actual task performance in realistic conditions.

Workload assessment methods include NASA-TLX (subjective), primary/secondary task measures, and physiological indices; no single measure captures all dimensions.

Procedure deviation is more likely when procedures conflict with actual plant state or are hard to follow; procedure design should support adaptation with clear boundaries.

Circadian low points typically occur around 02:00-06:00 and 14:00-16:00; alertness and performance are lowest during these windows.

Sleep debt accumulates with restricted sleep; recovery requires adequate sleep, and chronic restriction produces cumulative performance deficits.

Shift rotation direction matters: forward (day→evening→night) rotation is generally better tolerated than backward rotation because it aligns with circadian delay.

Fatigue risk factors include long duty hours, night work, insufficient rest between shifts, monotonous tasks, and high workload combined with low control.

Acute stress can narrow attention and increase reliance on habitual responses; chronic stress impairs memory, decision quality, and health.

Workload has multiple dimensions (mental, physical, temporal, effort, frustration, performance); high workload reduces spare capacity for unexpected events.

Countermeasures for fatigue include strategic napping, controlled rest, lighting, task rotation, and scheduling limits; caffeine is a temporary aid, not a substitute for sleep.

Team fatigue interacts: a fatigued operator may under-report and under-check; supervision should account for degraded self-monitoring.

Closed-loop communication: sender states message, receiver repeats back, sender confirms; this reduces misunderstanding and ensures shared understanding.

Standard phraseology and readback are core defences against communication error; ambiguity and non-standard terms increase error risk.

Briefings and debriefings align mental models, clarify roles, and surface concerns; pre-task briefs reduce coordination failures.

Assertiveness and psychological safety support speaking up; hierarchy and deference can suppress safety concerns (mitigation of authority gradient).

Shared mental models (task, team, and situation) improve coordination; they are built through briefings, cross-monitoring, and explicit updates.

Cross-monitoring (monitoring others' actions) provides an independent error-detection layer; it fails when workload is high or roles are unclear.

Team situational awareness requires information sharing; individual SA does not automatically produce team SA without communication.

Conflict and stress can degrade communication (tunnel vision, reduced information sharing); structured communication tools help maintain performance.

Safety culture dimensions commonly include management commitment, safety priority over production, learning, reporting, and just/fair culture.

Just culture distinguishes human error (console/support), at-risk behaviour (coach), and reckless behaviour (sanction); it encourages reporting without blame for honest error.

Reporting systems depend on trust and non-punitive response; under-reporting hides latent conditions and weakens learning.

Event investigation methods: root cause analysis (RCA), barrier analysis, and human performance investigation (HPI) focus on conditions and defences, not just the individual.

Accident causation models: Swiss cheese (Reason) shows defences breached by aligned holes; the model emphasizes latent conditions and organizational factors.

Human reliability analysis (HRA) quantifies error probability and identifies recovery; methods include THERP, HEART, and CREAM; results are estimates with uncertainty.

Performance indicators of safety culture include reporting rate, near-miss rate, audit findings closure, and safety climate survey results; lagging indicators alone are insufficient.

Investigation should separate what happened (sequence), why it happened (conditions), and what sustained it (organizational factors) to produce effective corrective actions.

Competence-based training uses defined performance standards, observable behaviours, and assessment against criteria rather than time served alone.

Training needs analysis (TNA) identifies task requirements, knowledge/skill gaps, and training objectives before designing content.

Simulator training supports scenario-based practice of abnormal and emergency events; fidelity should match the training objective (physical vs. functional fidelity).

Licensing and certification typically require initial training, examination, simulator assessment, and periodic revalidation with continuing training.

Crew resource management (CRM) training targets communication, teamwork, leadership, and decision making; it should be integrated with technical training.

Assessment reliability improves with multiple assessors, standardized scenarios, and behaviourally anchored rating scales; single-assessor judgments are less reliable.

Refresher and recurrent training counter skill decay; infrequently used emergency skills decay fastest and need more frequent practice.

Training records and competence matrices track qualification status; gaps between required and actual competence are a latent risk.