Human Reliability Assessment in Defence:
The Case for Real-Time Monitoring

Human Reliability Assessment (HRA) is a systematic methodology for quantifying the probability of human error in safety-critical operations. Embedded in NATO standards, MoD frameworks, and civilian nuclear and aviation safety protocols, HRA identifies the conditions under which human performance degrades — fatigue, cognitive overload, stress, time pressure — and assigns probability weights to failure modes. It is one of the most rigorous human factors methodologies in existence. It is also, by design, retrospective. EchoDepth applies the same dimensions HRA evaluates — in real time, continuously, without self-disclosure.

What HRA Measures and Why It Matters

HRA methodology traces to the nuclear industry, where quantifying the probability of human error in control room operations was a regulatory requirement from the 1970s. Techniques like THERP (Technique for Human Error Rate Prediction), HEART (Human Error Assessment and Reduction Technique), and ATHEANA (A Technique for Human Event Analysis) provide structured frameworks for estimating how frequently trained personnel will make specific errors under specified conditions.

In defence contexts, HRA is applied to UAS (drone) operations, submarine crews, artillery teams, intelligence analysis, and increasingly to cyber operations. Key performance-shaping factors in defence HRA include: fatigue state, cognitive load, stress level, training recency, team communication quality, and environmental stressors. These are the human variables that HRA frameworks require assessment professionals to evaluate.

The problem is the assessment mechanism. Traditional HRA uses structured interviews, expert judgement, and standardised questionnaires — all of which are administered at a single point in time. A pre-mission fitness assessment completed at 0700 does not capture the operator's state at 1400, four hours into a sustained UAS mission over a contested environment.

Fatigue: The Hidden Variable in Military Incident Causation

Research by Caldwell, Caldwell, and Darlington has established that operator fatigue is a contributory factor in approximately 21–23% of major military incident investigations. NATO review of aviation incidents consistently places fatigue in the top three human factors contributors to accidents and near-misses. Yet most operational environments have no real-time mechanism for detecting fatigue onset.

The standard approach is self-report: operators declare fatigue through Karolinska Sleepiness Scale ratings or equivalent instruments before missions. Self-report is unreliable for two reasons. First, operators in operational environments are subject to social pressure not to declare fatigue — reporting tiredness is perceived as weakness, particularly in high-performing military units. Second, fatigue impairs the cognitive capacity to accurately assess one's own fatigue state. Moderate fatigue reduces metacognitive accuracy — the ability to accurately self-monitor performance — making late-stage fatigue identification particularly unreliable.

The physiological signature of fatigue is clearly visible in facial Action Units before an operator's self-report would capture it: AU46 (eyelid droop) increases, micro-expression frequency decreases, and VAD arousal falls below the range associated with optimal alertness. EchoDepth captures these signals continuously, without requiring the operator to do anything differently.

Cognitive Load Monitoring in High-Stakes Operations

Beyond fatigue, cognitive overload is a distinct failure mode with a different facial signature. An operator managing excessive information throughput — multiple system alerts, radio communications, map updates — enters a state of cognitive overload characterised by elevated arousal, brow tension (AU4+AU7), and reduced attentional bandwidth.

Cognitive overload in SOC analysts is well-documented: alert fatigue — the desensitisation of analysts to security alerts due to excessive volume — is a primary cause of missed threat indicators. In control room operations, cognitive overload has been identified as a contributing factor in several major incidents including Three Mile Island (1979) and more recently in complex air traffic control failures.

EchoDepth's operator readiness monitoring capability provides a continuous cognitive load score alongside fatigue detection, enabling supervisors and automated alerting systems to identify when an operator's effective processing capacity is approaching critical thresholds — before an error occurs rather than after.

Extending HRA to the Real-Time Domain

The insight that connects HRA methodology to emotion recognition AI is straightforward: HRA frameworks evaluate the same dimensions that facial Action Unit analysis measures. Fatigue state, cognitive load, stress level, and arousal are not only theoretical constructs in HRA models — they have measurable, peer-reviewed physiological signatures in FACS AU patterns.

EchoDepth provides the evidence layer that HRA frameworks have always assumed but never been able to continuously capture. Rather than a pre-mission assessment that becomes stale within hours, EchoDepth delivers a continuous, timestamped record of the operator's state across the full mission — DSAT-compatible output that can support post-incident investigation, readiness certification, and training effectiveness assessment.

For teams operating under STANAG frameworks or JSP human factors requirements, EchoDepth provides the audit-ready, reproducible evidence layer that manual HRA processes cannot generate in real time. No wearables. No self-report dependency. No specialist hardware beyond an existing camera. SCIF and air-gap compatible deployment as standard.