How does an expert notice what a novice misses?

Two people can encounter the same situation and receive very different information from it. An experienced engineer hears a change in a machine. A doctor notices that a collection of ordinary symptoms does not fit the expected pattern. A pilot detects that an approach does not feel right. A musician hears a problem that most listeners cannot identify. The physical situation may be shared. What differs is the receiver. The Normal-Model Framework suggests that expertise changes what can be noticed, what appears meaningful and what is allowed to pass without further attention.

A richer model of normal

Experts develop detailed models of what normal looks, sounds and feels like within their domain. These models are not simply lists of facts. They contain relationships, patterns, boundaries, exceptions and expectations acquired through study and experience. A richer model of normal makes finer differences available. Where a novice sees one broad category, an expert may perceive several distinct possibilities. Where a novice sees a random collection of details, an expert may recognise a familiar configuration—or notice the one element that does not belong. Expertise therefore affects the resolution at which a situation can be represented. But detecting more differences is not enough.

Your normal is my difficult

Expertise can make the extraordinary look incidental. An expert may include a subtle distinction, adjustment or additional detail as though it costs nothing. What appears effortless now may compress years of learning and practice. It has become part of the expert’s normal, even though another receiver may struggle to produce or fully appreciate it. Recognising expertise also requires resolution. A non-expert may sense that a performance is unusual without understanding precisely what made it difficult or how well it was done. Performing, observing, diagnosing and teaching are related but different forms of expertise. A coach need not outperform the performer to possess expertise in perceiving and improving the performance. The expert’s throwaway detail may be the novice’s unreachable performance.

Resolution and discrimination

A person who treated every difference as important would quickly become overwhelmed. Expertise requires both: Resolution: the ability to detect finer differences; and Discrimination: the ability to judge which differences are likely to matter. An expert may notice more than a novice while investigating less of what they notice. They have learned which signals can usually be allowed to pass, which justify watching and which demand immediate attention. This ability to ignore well may be as important as the ability to notice. But it also creates vulnerability. A difference that falls outside the expert’s established categories may be dismissed because it resembles harmless variation. A strong model of normal can support recognition while also constraining it.

“Something isn’t right”

Expert detection may begin before the person can explain what they have noticed. A situation snags them: Wait a second … what’s this? They may initially have only a sense that something is not right. The relevant difference has entered attention, but its meaning has not yet been represented clearly enough to describe. This does not make the response magical or infallible. The feeling may reflect genuine expertise, an irrelevant association, anxiety, coincidence or noise. Its value lies not in proving that something is wrong, but in routing the situation towards further attention. The expert may then inspect, compare, test and reconstruct until the original snag becomes explainable—or is judged unimportant.

Diagnosis is a constructed representation

Diagnosis is often associated with medicine, but the same broad structure appears in many fields. A mechanic diagnoses a fault. An investigator diagnoses the cause of an inconsistency. A teacher diagnoses why a learner is struggling. A manager diagnoses an organisational problem. A technician diagnoses why a system has failed. In each case, diagnosis involves constructing a representation that makes the available signals meaningful. The diagnosis is not the underlying reality itself. It is a proposed account of that reality: a way of organising incomplete evidence sufficiently well to guide the next decision. A useful diagnosis may identify: • what is happening; • what may have caused it; • what evidence supports that interpretation; • what alternatives remain possible; • what further information is needed; and • what action is justified now.

The value and danger of a label

A diagnostic label compresses complexity. Once a pattern has a name, relevant knowledge can be retrieved, communicated and connected to possible action. The label allows people to coordinate without reconstructing the entire situation each time. But compression also removes detail. The label may conceal uncertainty, individual variation or competing explanations. Once accepted, it can reshape attention: evidence consistent with the diagnosis becomes easier to notice, while conflicting evidence may be treated as irrelevant. A diagnosis should therefore remain a representation that may be revised—not become confused with the reality it is intended to represent.

Expertise and diagnosis through the NMF

Detection What difference snagged the person? What could the expert detect because of their developed model of normal? What might remain invisible even to them? Evaluation What might the difference mean? The expert compares the signal with possible explanations, patterns and prior cases. They distinguish relevant evidence from noise and consider whether the situation fits an established category. Commitment When is the representation good enough to proceed? The expert may not need certainty. They need sufficient confidence for the next decision. The threshold required to conduct a harmless test may be lower than the threshold required for an irreversible intervention. Action What should happen next? Possible actions include testing, observing, consulting another person, treating, repairing, intervening, waiting or deliberately ruling out a possibility. Expertise includes matching the action to both the available confidence and the consequences of being wrong. Learning What should the outcome change? The outcome may strengthen, weaken or reorganise the expert’s model of normal. But feedback is not always clear. A correct diagnosis may be followed by a poor outcome. An incorrect diagnosis may appear successful because the problem resolves for another reason. Without careful learning, experience can reinforce the wrong model.

Experts recover sooner

Expertise is not only the ability to make a better initial judgment. It may also involve detecting sooner when the current representation is no longer working — and revising it before the mismatch becomes unmanageable. An expert acts with an expected result in mind. If Plan A does not produce that result, the difference between expectation and outcome becomes a new signal. The expert may then move through another DECAL sequence: Detection: A is not producing the expected result. Evaluation: What does this difference tell me? Commitment: B now appears to be the best available next step. Action: Try B. Learning: Compare what happens with what B was expected to produce. If B does not work, its outcome becomes input to another cycle. Plan B is not certain to succeed. It is the next defensible bet using the evidence and resources now available.

Adaptive lag

An expert cannot stay one step ahead of uncertainty. That would require knowing what will happen next. What they may be able to do is avoid falling too far behind it. Adaptive lag is the delay between reality no longer matching the receiver’s working representation and the receiver detecting, revising and responding to that mismatch. The situation may have changed, or the original representation may have been incomplete or mistaken. If the receiver adapts quickly enough, the situation may remain manageable. If events develop faster than the receiver can detect, evaluate and act, the consequences of the mismatch may escalate beyond their available cognitive or practical resources. The expert advantage may therefore include: • detecting earlier that expectation and reality are separating; • distinguishing meaningful failure from ordinary variation; • revising or abandoning an unsuccessful representation without waiting for catastrophic proof; • constructing plausible alternatives; • making the next defensible move; • preserving sufficient resources and flexibility to adjust again; and • completing another DECAL sequence before the situation gets out of hand. Good improvisation is not necessarily action without a model. It may be the rapid reconstruction of the model while acting. Uncertainty does not reveal its whole hand. Each outcome provides further evidence rather than complete truth. Adaptive expertise is therefore not the elimination of uncertainty. It is the capacity to detect early when reality no longer matches expectation, revise the working representation and make the next defensible move before the consequences become unmanageable.

Expertise is selective

Experts are sometimes described as people who know more. But expertise also concerns selection: • what they notice; • what they ignore; • what questions they ask; • what comparisons they make; • what units they use; • what possibilities they keep active; • when they seek more information; and • when they judge that they know enough to act. Different experts may construct different representations from the same evidence because they bring different models of normal, objectives and thresholds. Expertise does not eliminate perspective. It develops a perspective that has been refined for particular purposes.

Training the receiver

If expertise changes what can snag a person, training cannot be limited to transferring information. Training may need to develop the receiver’s model of normal. This can involve: • exposure to varied examples; • comparison of similar cases; • practice detecting subtle differences; • attention to exceptions and boundary cases; • explanation of which signals matter and why; • “what if?” variations; • feedback on both action and inaction; and • opportunities to revise a judgment before seeing the outcome. The learner needs examples of normal variation as well as examples of failure. Otherwise, they may become highly sensitive without becoming discriminating—detecting many differences but lacking the ability to judge which ones deserve attention.

Tools and distributed expertise

Experts rarely work alone. They use instruments, checklists, records, colleagues, procedures and increasingly artificial intelligence. These tools can extend resolution, preserve information and reveal patterns that an unaided person could not detect. Expertise may therefore reside partly in knowing: • which tool to use; • what question to ask; • how to interpret the result; • when the tool may be unreliable; and • when another person’s model is needed. The expert is not necessarily the person who personally knows everything. They may be the person who can construct an effective investigation from distributed sources of knowledge.

Questions NMF can bring to expertise and diagnosis

1. What model of normal is the expert using? 2. How was that model developed? 3. What can the expert detect that a novice cannot? 4. Which differences are treated as meaningful? 5. What is routinely allowed to pass? 6. What first snagged the person? 7. Could they explain why, or did explanation come later? 8. What alternative representations remain active? 9. What unit or comparison is being used? 10. What evidence would challenge the current diagnosis? 11. How much confidence is required for the next action? 12. Is the proposed action reversible? 13. What are the costs of acting, waiting or investigating further? 14. What will the outcome genuinely allow the person to learn? 15. Could experience be reinforcing the wrong lesson? 16. How quickly was the mismatch between expectation and outcome detected? 17. What justified abandoning or revising the current representation? 18. What alternative actions remained available? 19. Had sufficient resources and flexibility been preserved for another attempt? 20. How much adaptive lag could the situation tolerate before it became unmanageable?

A potential use — not a diagnostic method

The Normal-Model Framework does not diagnose medical conditions, identify mechanical faults or determine who qualifies as an expert. It does not replace specialist knowledge, professional standards, evidence-based diagnostic methods or second opinions. It offers a way to examine how expertise changes noticing, how a difference becomes diagnostically meaningful, how confidence reaches a threshold and how outcomes reshape future judgment. Its potential contribution lies in making expert recognition more open to examination: What did the expert notice? What made it meaningful? Why was that representation judged good enough? And what should happen to the model if the outcome is unexpected?