Wednesday morning, March 11th. A CEO dismisses a stack of hiring data with a wave of her hand. "I don't need the numbers. After twenty years in this industry, I know a great candidate when I see one." She believes this completely. She is partially right and mostly wrong. Her certainty is the problem.

Intuition is real. It's a genuine cognitive phenomenon grounded in compressed experience and pattern recognition. It also fails systematically in most of the high-stakes domains where people invoke it with the most confidence. The advice to "trust your gut" is sound in exactly the contexts where nobody needs to be told to trust their gut — and catastrophically bad advice in the contexts where people most want to hear it.

What Intuition Actually Is

Intuition is not mystical. It's pattern recognition operating below the threshold of conscious deliberation. Chess grandmasters don't calculate twenty moves ahead for every position — they glance at the board and immediately recognize which patterns are relevant. Experienced firefighters enter burning buildings and sense danger before they can articulate why. Nurses develop rapid reads on deteriorating patients before the monitors confirm it.

Psychologist Gary Klein spent decades studying this phenomenon through what he called naturalistic decision-making: studying how experts make decisions under real conditions, with time pressure and incomplete information. His conclusion: expert intuition is real, it's powerful, and it works through rapid pattern matching to an extensive library of recognized situations built through experience.

The research is clear on the mechanism. Intuition works when: you've had extensive experience in a domain, that domain provides rapid and accurate feedback on whether your judgments were right, and the underlying patterns in the domain are stable enough to actually learn from.

This is not most situations. And it's not the situations people are usually talking about when they say "trust your gut."

When the Signal Breaks

Psychologist Paul Meehl published a quietly devastating paper in 1954: clinical psychologists predicting patient outcomes performed worse, on average, than simple statistical formulas using the same information. In a 1996 review, William Grove found this result held across 136 comparisons of clinical versus actuarial judgment. Experienced clinicians, relying on intuition refined over years of practice, consistently lost to the algorithm.

This is not an isolated finding. Philip Tetlock's landmark study, tracking 28,000 expert predictions on political and economic questions over twenty years, found that experts did only slightly better than chance at predicting outside their narrow specialty — and expressed approximately the same confidence in their wrong predictions as in their right ones. The feedback loop for "I think this political situation will unfold this way" is too slow and too noisy to ever build genuine pattern recognition. The expert's gut accumulated not calibrated judgment but accumulated confidence.

Kahneman and Klein collaborated on a paper in 2009 asking directly: when can you trust expert intuition? Their joint answer was that intuitive expertise requires what they called a "high-validity environment" — one where cues reliably predict outcomes, and where the feedback on your predictions is clear and timely. Chess: yes. Firefighting: yes. Long-term stock picking: no. Hiring: no. Clinical prognosis: mixed. Strategic planning: usually no.

The problem with domains like hiring, investment, and diagnosis is not that intuition isn't trying. It's that the feedback never arrives in a form that trains the signal. You hire someone who seems great — three years later you still can't tell if they'd have failed under a different manager, or if the one who seemed less promising would have turned out equally well. The pattern library builds, but it builds on noise. Confidence accrues without calibration.

The Confidence Gap

Here is the specific mechanism by which this causes damage: confidence and accuracy are correlated in high-feedback domains and essentially uncorrelated in low-feedback domains.

A grandmaster who feels certain about a position is probably right. Certainty, there, has been calibrated by millions of games with clear outcomes. An investor who feels certain about a macro call is probably just confident. The certainty is the same feeling. The reliability is completely different.

Worse: in low-feedback domains, the people with the most experience often have the most confidence without any corresponding improvement in accuracy. They've practiced longer, which means they've built a more extensive and fluent library of patterns — patterns that don't actually predict outcomes. The experience is real. The expertise is an illusion of the experience.

This is what makes the CEO so dangerous. She's built something real over twenty years — a rich, nuanced model of what a great candidate looks like. That model is based on thousands of interviews, and she has received feedback on some of those hires. But the feedback is confounded at every step: her treatment of the hire influenced the outcome, the team influenced the outcome, market conditions influenced the outcome. She can't cleanly separate "my gut was right about this person" from "this context was good for almost everyone." The signal she thinks she's been training on is a noise field.

The Systematic Damage

Unwarranted faith in gut judgment causes specific, predictable errors.

In hiring, it systematically advantages candidates who fit preexisting patterns — who look, sound, and present themselves like people who succeeded in the past. This isn't conscious bias; it's the natural output of pattern matching to a biased training set. Structured interviews with scored rubrics consistently outpredict "impression-based" hiring. Most organizations don't use them because the hiring manager's gut disagrees with the rubric.

In strategy, it produces overconfidence in inside-view reasoning — "I know this market, I know this customer, I know what will work." The inside view is the gut's native domain. The outside view (what's the base rate? what usually happens to people who try this?) requires deliberate effort precisely because it's not what intuition returns. The gut story feels richer and more vivid than the base rate, even when the base rate is more accurate.

In risk assessment, expert intuition tends to compress low-probability catastrophic risks (they feel implausible, the pattern library contains few of them) while amplifying vivid recent dangers (the availability heuristic encodes them as common). The experts with the most experience making these judgments are not more accurate — they're more fluent in an unreliable process.

The Narrow Zone Where It Works

None of this means you should ignore gut reactions. It means you should take them seriously in the right contexts and discount them in others.

Trust intuition when:

  • You are operating in a domain you've practiced extensively, with rapid feedback that told you whether you were right
  • You are recognizing a type of situation you've encountered before under conditions where the feedback loop was clean
  • The decision is under time pressure and the cost of a slower analytical process exceeds the cost of error

Be skeptical of intuition when:

  • The domain has slow, noisy, or confounded feedback
  • The stakes are high enough that calibration error matters
  • You can articulate a specific analytical alternative that doesn't just confirm the gut reading

The tell for "this is high-validity intuition" is not how strong the feeling is. It's whether the feeling comes with pattern recognition you can almost articulate — "this reminds me of X situation, and X situations tend to end badly because of Y." Vague certainty is not the same thing.

Takeaways

The gut is a pattern-matching system. Like any pattern-matcher, its output quality is entirely determined by the quality of the patterns it learned from. In domains with tight feedback loops and stable regularities, it's a powerful tool. In domains without those properties, it's a confidence machine that generates strong feelings unmoored from accuracy.

Concrete adjustments:

  1. Map your domains. For each major class of decision you make repeatedly, ask: how fast does feedback arrive? How cleanly can I trace outcomes to my judgment versus other factors? Low-validity domains are where gut instinct reliably fails.

  2. Use structured methods in low-validity domains. Not because structured methods are perfect, but because they're less exposed to the specific failure mode of gut judgment: systematically training on noise and building fluent but uncalibrated pattern libraries. Rubrics, checklists, and base rates are not bureaucracy — they're corrections for a predictable error.

  3. Distinguish recognition from certainty. The valuable output of expert intuition is "I've seen something like this before." That recognition is worth taking seriously as a hypothesis to investigate. The leap to "therefore I'm right" is the trap. Recognition is evidence. Certainty is not.

  4. Audit past gut calls in low-validity domains. Pick twenty decisions from the last few years where you trusted your instinct. How many panned out? How many did you attribute to other factors when they didn't? The calibration exercise is uncomfortable and useful.

  5. Be especially suspicious of confident gut calls on irreversible decisions. The domains with the poorest feedback loops — hiring, partnerships, strategy — are also the domains where the decisions are hardest to reverse. The combination of low calibration and high irreversibility is where the damage compounds.

The goal is not to suppress intuition but to route it correctly. In a high-feedback domain, your gut has been built by reality and is worth listening to. In a low-feedback domain, it's been built by whatever you happened to notice — which is not the same thing.

The grandmaster's gut sees the board. The executive's gut sees themselves.

Today's Sketch

March 11, 2026