Human Factors in Accident Investigation: Why Tacit Expert Knowledge Is A Critical Missing Piece

When I am involved in a serious accident investigation, I run a thought experiment with the client early on. I ask them to identify the top resident expert for the task involved – let’s call this person Joe. Then I ask: if Joe had been doing this task on that night shift, do you think this accident would reasonably have happened?

Almost without exception, the answer is: no. Absolutely not!

So what is the difference between Joe and the two engineers on that shift — and what does that tell us about the root causes of accidents? Clearly, it is the tacit knowledge, the skills, experience, or more precisely, expertise. It is this resident expertise – what can be thought of as cognitive diamonds – that ensures critical work operations are performed safely 99.99% of the time.

Cognitive diamonds fill the gap between standard operating procedures and the ever-shifting, adaptive reality on the ground. They are what completes the work design system. The presence or absence of these cognitive diamonds is what so often makes the difference between safe and unsafe completion of a critical task.

But what exactly are cognitive diamonds? The best way to answer that is to start with Gary Klein’s work, a leading researcher in the field of expert decision-making in naturalistic workplace situations.

Klein developed his Recognition-Primed Decision-making (RPD) model through fieldwork with experts in high-stakes environments, including firefighters (Klein, 1993). What he found puzzled him at first. When he asked experienced firefighters how they made critical, often life-saving decisions under pressure, they said – “We don’t make decisions, we just know.”

They described it as a sixth sense – an ability to size up a situation and act correctly, instantly, in ways they couldn’t explain. To their colleagues, it looked like a magical superpower.

Klein describes one incident that captures this perfectly. An experienced fire commander led his team into a burning building. He took one look and shouted, “Everyone out. Now.” The team got out. Seconds later, the building collapsed.

How did he know? Through a cognitive investigation technique called Cognitive Task Analysis (CTA), Klein’s interviews unpacked in detail how the fire ground commander had perceived, understood, and made his decision, in what had felt like the blink of an eye. The commander had entered what he assumed was the first floor.

Something felt deeply wrong: intense heat, very little visible fire or smoke. He had registered it as a timber-frame building on the side of a hill. Putting these cues together instantaneously, his decision came as an automatic response: the fire is below us, it is large and well-advanced, and in a timber-frame building on a hillside, that floor is going to collapse very soon. Everyone out.

That pattern – what to look for, what it means, and how to act – is a cognitive diamond. It operates across all three stages of situational awareness (Endsley, 1995): perceiving the right cues (Stage 1), understanding what they mean (Stage 2), and making the right decision (Stage 3). Such tacit knowledge is built through years of experience and operates below the level of conscious awareness – which is why experts struggle to articulate it.

Cognitive diamonds provide what researchers call decision heuristics – automatic processes for making effective decisions in complex, adaptive situations. Research by Reb et al. (2024) has shown how decision heuristics, once surfaced through CTA, support highly complex decision-making in adaptive and critical situations, functioning as flexible mental algorithms that cope with the unforeseen in ways that rigid procedures cannot.

Just as Klein did with firefighters, I have been able to unpack the cognitive diamonds of experts across a wide range of workplaces where serious incidents have occurred – including the maintenance engineers at the food manufacturing plant at the centre of this research. These cognitive diamonds are rarely, if ever, documented anywhere. They are not in the standard operating procedures. They are not in the training manual. They live entirely inside the heads of resident experts – invisible to conventional investigation and invisible to the organisation until specifically and deliberately surfaced. They are what deliver successful performance. And as we will see, they can be unpacked and leveraged – to accelerate the development of expertise and to design tasks that better support decision-making in critical situations.

Cognitive task analysis (CTA) is a key methodology designed to surface tacit knowledge or expertise. The specific technique most applicable to accident investigation is the Critical Decision Method (CDM), developed by Klein et al. (1989) and applied in my research (Pemberton & Mooney, 2023).

The CDM starts by building a timeline of the event with the participant – a first pass maps the sequence of actions and decisions from start to finish. The interviewer then makes multiple further passes through that timeline, each going deeper into the cognitive work at the critical decision points: the moments where different choices could have led to different outcomes.

At each decision point, structured cognitive probes surface the underlying mental work: What were you attending to here? What cues told you something was right or wrong? What did you know that a less experienced person might not have known? What would an expert have seen or done differently? What would have needed to be different for you to have acted differently?

Crucially, to turbocharge learning from accidents, CTA interviews should be conducted with two groups – not just the people directly involved, but also resident experts who were not involved.

My research demonstrates that the richest learning comes from the expert simulation interviews. The expert is taken through a simulation interview that walks them through the accident scenario to explore what they would have perceived, understood, and decided at each critical point. That account is then compared directly with what the accident participants knew, understood and decided. The gap between the two is where the most important learning lives.

“CTA, by unearthing cognitive diamonds, is aiming to help develop decision heuristics – the flexible mental algorithms that kick in precisely where the procedures end and adaptive reality begins (Reb et al., 2024). The SOP describes the recipe. What the expert actually does to achieve that unbelievable level of safe performance – that if only it could be passed on to the rest of their colleagues would make a major difference to safety performance – is simply not being captured.

What’s preventing this deeper learning is simply that the large majority of conventionally trained accident investigators are not equipped to navigate this territory (Pemberton & Mooney, 2023).”

Implementing CTA well requires significantly higher skill than conventional accident investigation interviews. For example, when taking a resident expert through a simulation interview, they will almost always default initially to the standard operating procedure: “We have an SOP for this, which requires…” That is exactly the moment where the CTA interviewer’s skill is critical – pull them back.

Not what the SOP says to – “What would you actually do? What would you look for that is not in the procedure? What would tell you something was wrong?” It is in navigating that gap that the cognitive diamonds are found.

The structured output of the CTA process is a Decision Requirements Table (DRT). For each difficult decision – where a less experienced practitioner would be likely to get it wrong – the DRT documents the cognitive demands, the cues experts use to read the situation, the mental models that underpin expert judgement. The DRT is the bridge between investigation insight and practical action: the data needed for simulation training design and for improving work system and task design.

Back to the night shift at the food manufacturing plant.

The task involved replacing a valve on a pressurised system. In the expert simulation interviews, resident engineers described a rich set of cues they would attend to before starting work. “I would look at the floor around the work area: fresh drip marks mean a new leak; old staining means a long-standing one.”

But what do these cues mean? “A long-standing leak changes the risk picture entirely – indicating possible corrosion or system degradation that fundamentally affects how the job should be approached. Spanner marks on the nut tell me whether the fitting has been recently worked and whether it is likely to be brittle or stubborn.”

The big picture view is another aspect that sharply differentiates experts from less experienced workers – the ability to see the task in its wider operational context. It was precisely this lack of big picture thinking that contributed to catching out the engineers in this case.

Resident experts understood that even with the system isolated, the tank above could still contain a significant volume of very hot liquid. They were simultaneously attending to the specific details of the valve replacement and to the wider process context. Just as Klein’s fire commander was reading heat, smoke, and building type, he was able to see that these cues meant the building was about to collapse. Situational awareness operating at multiple levels at once.

The two engineers on that night shift were focused on the valve replacement. The wider process picture – and the risk it contained – was not in their mental model. Not negligence. Not carelessness. This is the nuance that is often missed when discussing human error in workplace accidents – they simply did not yet have the cognitive diamonds.

In this example, the CTA and subsequent DRT uncovered seven difficult decisions – three of which involved non-technical safety skills (NoTSS). NoTSS are the cognitive and social skills that underpin safe performance beyond technical knowledge: situational awareness, communication, teamwork, decision-making under pressure, and the ability to assertively raise concerns (Flin et al., 2008). The most revealing of the three NoTSS-based decisions was how to negotiate a delay when a problem is spotted under production pressure.

This is where it gets very human. The research showed the resident experts were able to push back to avoid what they described as playing the safety card – walking the supervisor through exactly what they had seen, what it meant, and what could happen if the job proceeded.

Supervisors hear that all the time. What the resident experts had, and the engineers involved lacked, was the ability to take someone through the big picture in detail: “here is what I found, here is what it tells me about the system state, here is what could happen if we proceed and things don’t go to plan.” That kind of structured, evidence-based assertiveness is a non-technical safety skill – and it is every bit as safety-critical as knowing which spanner to use (Pemberton & Mooney, 2023; Flin et al., 2008).

The cognitive diamonds and decision heuristics documented in the DRT are directly actionable as training content. One practical and highly effective form of delivering that training is through Decision-Making Exercises (DMXs).

A DMX is a form of simulation training that places participants in realistic dilemma situations – recreations of the kinds of complex, ambiguous decisions that characterise their real work. The scenario develops in stages and at key decision points, participants work through what they perceive, what they understand, and what they would do.

Short and highly interactive, DMXs are built directly from the difficult decisions and cognitive diamonds in the DRT – not generic training, but training built from the tacit knowledge of your own resident experts, targeted at the specific cognitive demands of your own critical tasks. DMXs range from paper-and-pencil exercises to sophisticated online, AI-supported simulation platforms, and work equally well for technical cognitive skills and non-technical safety skills.

They generate real engagement. I regularly see operational teams walking out of DMX sessions still arguing over the decisions they faced. That is exactly what you want: genuine cognitive work, in a safe environment, on the decisions that matter most.

Consider the knowledge gap in my research. I benchmarked experienced workers – averaging nearly ten years of operational experience – against the cognitive diamonds documented by resident experts (Pemberton & Mooney, 2023). So they were not new starters – seasoned operational employees. And yet they demonstrated less than a quarter of the resident expert tacit knowledge. Left to chance, tacit knowledge transfer is slow, inconsistent, and deeply incomplete.

In the maintenance task, resident experts identified over 14 factors for pre-job planning – for example, specialist welding support if the union broke off, scaffolding arrangements, specialist lifting equipment, impact on the wider plant if the job overran, and more. The engineers involved had considered two or three. Not a reflection of their commitment – a reflection of how much cognitive expertise goes undocumented and untransferred in most organisations.

DMX training built from the DRT produced significant improvements in tacit knowledge – across both technical and non-technical safety skills – in little more than two hours (Pemberton & Mooney, 2023). Structured cognitive skill development can dramatically compress the time normally required to reach expert-level performance (Hoffman et al., 2014; Militello, 2023).

Training is one application of cognitive diamonds and decision heuristics. The DRT insights can also be used to redesign the work system itself – so safe decisions become more consistent regardless of an individual worker’s current level of expertise.

In the maintenance task, mapping where the difficult decisions occur makes it possible to design task support tools that embed resident expert cognitive diamonds directly into the work process. Decision support prompts in job planning – flagging when to pause, what to check, and what specialist support to have in place.

Pre-job planning tools can be built from DRTs that go beyond generic risk assessment templates. AI-based tools are also increasingly being used to encapsulate expert knowledge and guide teams through structured pre-work conversations, making the cognitive diamonds of the best practitioners available to everyone, every shift.

This is about improving what is sometimes called the joint cognitive system – the partnership between people, tools, environment, and processes that determines how well humans and their work systems function together. The goal is to design the work system so that expert-level thinking is easier to apply. And when things do start to go wrong, which they will, the system actively supports workers in catching errors so the system can fail safely.

In my experience, cognitive diamonds are rarely explored in accident investigation – or in any other safety initiative, for that matter. As a result, they are largely undocumented or used to improve safety performance. Their huge benefits are restricted to the small number of resident experts in whose heads they reside.

Training is therefore restricted to recipe following: procedures, checklists, and standard instructions. The tacit job smarts – the real cognitive expertise that makes critical tasks go right the vast majority of the time – are acquired on the job, passed on through painful trial and error or ad hoc coaching by more experienced workers. It’s hit-and-miss, slow, and entirely dependent on the availability of experienced practitioners willing and able to share what they know.

This model is no longer viable. Across the organisations I work with, average tenure is dropping significantly. The era of the long-serving resident expert with decades of deep operational knowledge is declining. As experienced practitioners leave or retire, their cognitive diamonds leave with them – quietly, undramatically, and with enormous consequences for safety performance. This is intellectual property that literally runs the organisation – the accumulated cognitive expertise that makes critical operations successful, that keeps people safe, that differentiates high-performing teams from average ones. Is there anything more valuable? And yet in most organisations, it receives almost no systematic attention.

Operating procedures are the recipe. A recipe does not make a chef. In critical tasks involving complex, adaptive situations – where it is impossible to pre-plan everything in an SOP – you need chefs. And the pipeline of chefs is shrinking.

Klein (1992) identified this risk decades ago, and crucially developed the methodology and technology – cognitive task analysis – to address it: a practical mechanism for capturing and preserving corporate memory before it is lost. The cognitive diamonds held by resident experts are not just a safety asset – they are an operational asset, critical to quality, efficiency, and overall performance. Their loss is a form of organisational amnesia that conventional training and knowledge management cannot address.

A cognitive human factors approach – CTA specifically – provides an evidence-based, engineered way of unpacking, documenting, and leveraging resident expert tacit knowledge for safety performance improvement. It captures the diamonds before they walk out the door. And through DMX simulation training, task design improvement, and cognitive system redesign, it transfers and embeds that expertise across the organisation – at a fraction of the time unstructured experience alone would require (Hoffman et al., 2014; Militello, 2023).

Genuine human factors in accident investigation means going beyond just the physical to also explore the cognitive dimension: how people were thinking, deciding, and reading the situation at critical moments. That cognitive dimension is where hugely important lessons from serious accidents live. And, due to a routine lack of accident investigator competency, almost entirely invisible to conventional investigations.

Cognitive diamonds are the tacit decision heuristics that expert practitioners use to navigate complex, adaptive work situations safely. They explain how resident experts can more often avoid serious accidents. Cognitive diamonds are what make critical tasks go right the vast majority of the time. They are built through years of experience, they operate below conscious awareness, and they are rarely, if ever, documented.

Getting better learning from accident investigations requires surfacing them. The CTA Critical Decision Method does exactly that – through structured interviews with both accident participants and resident experts, making the invisible visible.

The Decision Requirements Table (DRT) documents resident expert tacit knowledge for wider safety use: for training design, task improvement, and work system redesign. Decision-making exercises (DMX) are practical simulation training sessions that can transfer tacit knowledge rapidly and measurably to less experienced workers.

Through DMXs tacit knowledge development can be accelerated – from decades to months. And cognitive task design embeds this knowledge into work systems, so expert-level decision-making becomes easier.

This is what human factors in accident investigation can deliver. Not just a better understanding of what went wrong – but a systematic capture of what needs to go right, and a practical plan for making it happen.

At Human Focus International, we provide specialist Human Factors in Accident Investigation training covering the full CTA approach described here – including the CTA Critical Decision Method, resident expert simulation interviews, difficult decision identification, and building Decision Requirements Tables that capture the cognitive diamonds needed for training and task design.

Our services extend to DMX simulation training built directly from resident expert cognitive diamonds, consultancy and coaching for cognitive task design and joint cognitive system improvement, and an online competency development system that makes evidence-based cognitive skills training accessible at scale.

If you want to start capturing the cognitive diamonds in your organisation before they are lost – we would welcome a conversation.

Contact Human Focus International | Download the Research Paper: Pemberton & Mooney, 2023

Ian Pemberton is a Chartered Ergonomist with over 35 years of professional practice, specialising in human factors, cognitive task analysis, and accident investigation. He is the founder of Human Focus International.

This blog is based on Pemberton, I. & Mooney, D. (2023). How CTA improves learning from serious accidents. Human Focus International.