The Role Of Intuition In Expert Judgment

Gary Klein's research program began in the 1980s when the U.S. Army asked him to study how experienced commanders make decisions under field conditions. The classical model of rational decision-making — enumerate options, evaluate each against criteria, choose the best — seemed like the obviously correct framework. Klein went in expecting to document it.

He found something completely different.

Experienced decision-makers under time pressure and uncertainty didn't enumerate options. They generated a single candidate option based on pattern recognition and then ran a rapid mental simulation to check if it would work. If yes, execute. If no, modify or reject and generate the next candidate — still one at a time. They were running sequential single-option evaluation, not comparative multi-option analysis.

The reason they could do this: their experience had compressed situation assessment into pattern recognition. When a commander sees a situation that resembles situations they've encountered before, they immediately and automatically have a sense of what's happening and what to do. The pattern recognition produces an action option. They then use deliberate cognition to check that option quickly rather than to generate it from scratch.

Klein documented this across domains: firefighters, intensive care nurses, chess players, military planners, design engineers, critical care flight nurses. The RPD model held consistently.

The famous case is Lieutenant Commander Randal Comer, a nuclear power plant operator Klein interviewed. In a training scenario, Comer began executing a shutdown of a pump even though diagnostics showed the pump was fine. He couldn't explain why. Post-hoc analysis showed he was responding to a subtle asymmetry in temperature readings that suggested an incipient pump problem — a pattern he'd encountered before and encoded implicitly. His diagnosis was correct. But he couldn't articulate it.

This is the phenomenology of expert intuition: a judgment that arrives without the visible chain of reasoning, that is nonetheless accurate, and that is experienced as a felt sense or confidence rather than as a conclusion of deliberation.

Kahneman and Klein's joint paper — "Conditions for Intuitive Expertise: A Failure to Disagree," published in American Psychologist in 2009 — represents a genuine convergence between researchers who spent decades on opposite sides of the intuition debate. Their synthesis identified the conditions under which intuition can be trusted:

1. Regularity in the environment.

The environment must have stable underlying patterns that can be learned. Chess is a paradigm case: the rules are fixed, positions repeat with variations, the consequences of moves are deterministic. The environment is highly regular.

Financial markets are a counterexample. Market dynamics are produced by human behavior that itself responds to information about market dynamics — creating reflexivity. Patterns that held during one regime don't hold in another. An investor's intuitions calibrated in one market environment may be positively misleading in a structurally different one.

Medicine falls somewhere in the middle. Physiological processes are regular enough for expert intuition to be valuable — the pattern of symptoms, the presentation of specific conditions. But medicine also involves high variability in individual patients, complex interactions, and delayed feedback that compromises calibration.

2. Extended practice with that environment.

Pattern recognition requires exposure to many instances. The estimate of 10,000 hours (from Ericsson's deliberate practice research) is approximate, but the underlying point is robust: expertise requires large numbers of encounters with the domain, not just a few.

Critically, this must be deliberate practice — exposure that is actively processed, that involves effort to understand and encode, that is reflected on. Passive exposure doesn't build pattern recognition at the same rate. A surgeon who has performed 5,000 procedures while thinking carefully about each one has built different (and better) pattern recognition than one who has performed 5,000 procedures while on autopilot.

3. Reliable and timely feedback.

This is where many putative "expert" domains fail. If feedback on your judgments is delayed, rare, or ambiguous, your pattern recognition can't calibrate. You'll develop confidence without accuracy.

Clinical psychologists making long-term patient prognoses receive feedback on a small fraction of their predictions, long after the prediction was made, in a form they can often attribute to other causes. This is why clinical judgment in that domain is often poorly calibrated. The feedback loop is broken.

Radiologists reading scans receive rapid, concrete feedback when biopsies confirm or disconfirm their reads. Their pattern recognition can calibrate. Studies consistently show radiologists are better-calibrated than clinicians who receive slower, noisier feedback.

The implication: an expert in a low-feedback domain may have high confidence and low accuracy simultaneously. They're not lying or stupid — they've simply been unable to calibrate because the environment didn't give them the information they needed.

Klein's work established that expert intuition is real. Kahneman's work established that it fails in specific, predictable ways. The overlap is where the practical learning lives.

Corrupted feedback. When outcomes are determined by many factors besides your judgment, you can't learn from them reliably. Investment returns are affected by market conditions, luck, timing, and many decisions other than the one you're evaluating. Medical outcomes are affected by patient behavior, random factors, and co-treatments. In these environments, "I've been doing this for 20 years" is not strong evidence of well-calibrated intuition.

Low base rates. If the event you're judging is rare, you may never have encountered enough instances to build real pattern recognition. A judge ruling on an unusual crime type. A doctor seeing a rare disease presentation. An investor evaluating an unprecedented business model. The expertise that works for common cases doesn't transfer cleanly to rare ones.

Emotional contamination. Strong emotions corrupt pattern recognition by making certain patterns more salient regardless of their actual validity. A doctor who has recently seen a dramatic case of X will be more likely to see X in the next ambiguous presentation. A manager who has recently hired a successful employee from a certain background will overweight that background signal. The recency and emotional charge override the actual base rate information.

Structural change in the environment. The most dangerous failure mode. If the underlying structure of a domain changes — a technology shift, a regulatory change, a market structure shift — previously valid patterns become invalid. The intuition continues to feel valid because it's built from real experience. The experience is just no longer relevant.

Transfer to adjacent domains. Expertise in one domain produces overconfidence in adjacent domains. A successful entrepreneur whose pattern recognition is built in one industry will feel confident in another industry — and sometimes will be right, if the domains are genuinely similar. Often they won't be. The expertise didn't transfer; the confidence did.

What does the convergence between these two research programs mean for how you use your intuition?

Trust your gut when:

- You have substantial, deliberate experience in this specific type of situation - The environment you built that experience in has the same basic structure as the current situation - You've received reliable feedback on similar judgments and know you're well-calibrated here - Time pressure or information limits make deliberate analysis impossible anyway

Override your gut when:

- The situation is structurally novel — not just unfamiliar but genuinely different from what you've encountered - You're operating in a domain where feedback is long, rare, or corrupted - You can articulate a reason to be suspicious of your gut feeling (recent emotional event, stakes that might trigger motivated reasoning, etc.) - Data is available and contradicts your intuition — not always, but take it seriously

The discipline to develop:

Stop treating all confident feelings as valid signals. Confidence is generated by the same process that generates intuition — it doesn't distinguish between well-calibrated pattern recognition and overconfident pattern-matching on a domain you don't actually know.

The expert who says "I just have a feeling about this" in their genuine domain is worth listening to. The same expert saying the same thing in a domain they don't know is no more reliable than a novice — and possibly less, because the novice knows they don't know.

If expert intuition is accumulated pattern recognition from experience plus feedback, then you can engineer the conditions to build it faster.

Increase deliberate exposure. Volume of experience matters. The person who deliberately puts themselves in more situations of the relevant type — more decisions, more negotiations, more diagnostic cases — builds pattern recognition faster than the person with less exposure.

Create your own feedback loops. Where the environment doesn't provide feedback quickly, create it. A decision journal (see law_2_086) is essentially a tool for creating feedback loops in domains where natural feedback is delayed. If you make predictions explicit and then review them, you get the feedback even when the natural environment doesn't deliver it clearly.

Study patterns explicitly. Don't just do the work — analyze it. After a negotiation, review what happened. After a project, do a post-mortem. After a clinical case, read the outcome notes. Deliberate practice means actively processing experience, not just accumulating it.

Find a domain with fast feedback cycles. If you have a choice of domains to develop expertise in, domains with faster and clearer feedback will build better-calibrated intuition more quickly. Trading, writing, design, coding — all provide relatively rapid feedback. Management, policy, investing — feedback is slower and noisier.

Expose yourself to your failure cases. Expert intuition calibrated only on successes will overestimate its own accuracy. Deliberately seek out the cases where you were wrong. Study them. Let them inform and constrain your pattern recognition.

The person who does this work for a decade has something genuinely valuable: intuition they can partly trust, in domains they know, with enough self-knowledge to recognize when they're operating outside those domains. That combination — calibrated intuition plus the wisdom to know its limits — is what expert judgment actually looks like from the inside.

It's not magic. It's accumulated work made fast.