The Failure to Disagree: How Two Rival Schools of Decision Science Found Common Ground

Opening: A fight that never quite happened

Academic rivalries usually announce themselves loudly. Two camps stake out opposing positions, trade papers like artillery fire, and the field spends a decade picking sides. Something stranger happened in 2009, when Daniel Kahneman, the psychologist whose work on heuristics and biases had spent thirty years cataloging the ways human judgment goes wrong, sat down to write a joint paper with Gary Klein, a psychologist whose entire career had been built on documenting how firefighters, nurses, and battlefield commanders get it remarkably right using nothing but gut instinct. Two scholars, two decades of talking past each other, one paper. Its title told you everything before you read a word: "Conditions for Intuitive Expertise: A Failure to Disagree."¹

It's a strange kind of headline for an academic paper, an admission of failure dressed up as a finding. But it captured something true. Kahneman and Klein had set out expecting to referee a fight between their two schools of thought, and instead discovered that once they specified their terms carefully enough, there wasn't much of a fight left to referee.² That discovery, and everything it implies for how the rest of us think about expertise, intuition, and improvisation, is the real subject of this piece. But to understand why the paper mattered, you first have to understand how far apart the two camps had drifted.

Two schools, two very different starting points

The Heuristics and Biases program, born out of Kahneman and Amos Tversky's work in the 1970s, starts from a discomforting premise: our intuitions are frequently, systematically, and predictably wrong. Representativeness, availability, anchoring, these aren't random noise in human judgment, they're stable shortcuts the mind uses that produce stable errors, the kind you can replicate in a lab with strangers who've never met each other. The tradition's signature move is the puzzle that fools nearly everyone: the bat-and-ball problem, the Linda the feminist bank teller problem, the studies showing that experienced radiologists disagree with themselves on the same X-ray depending on the day. If human judgment fails this reliably under controlled conditions, the reasoning goes, we should be deeply skeptical of intuition wherever the stakes are higher and the conditions less controlled.

Naturalistic Decision Making, developed largely by Gary Klein starting in the mid-1980s, starts from the opposite intuition, quite literally. Klein didn't build his theory in a lab with undergraduates and gambles. He built it by watching actual fireground commanders make actual decisions, and what he found broke the standard model of decision-making almost immediately. In the vast majority of critical incidents Klein studied, commanders didn't compare multiple options at all. They looked at a fire, recognized it as a variation of something they'd seen before, and moved straight to action, refining the plan through mental simulation rather than deliberation among alternatives.³⁴ Klein called this the recognition-primed decision model, RPD for short, and it became the foundational finding of the entire NDM tradition.⁵⁴

For roughly two decades, these two traditions developed almost in parallel, citing each other mostly to disagree. Gigerenzer, working alongside Klein's broader intellectual camp, argued that Kahneman's biases were often just simple heuristics being ecologically rational in the real world, cheap rules that exploit the actual structure of an environment rather than deviating from some abstract statistical ideal. Kahneman and Tversky pushed back sharply in their 1996 paper "On the Reality of Cognitive Illusions," insisting the errors they'd documented were robust even under the ecologically-friendly conditions Gigerenzer favored. It had the shape of a genuine rift: one side insisted intuition needed correcting, the other insisted intuition, properly trained, needed protecting.

The question that forced them together

What finally brought Kahneman and Klein into the same room wasn't a truce negotiated by a third party. It was a shared recognition that the argument, as usually framed, was unanswerable because both sides were talking about different kinds of intuition without saying so. Klein's fireground commanders had spent years, sometimes decades, in an environment with fast, clear feedback: a fire either behaves the way you predicted or it doesn't, and you find out within minutes. Kahneman's overconfident forecasters, stock pickers, and radiologists often operated in environments where feedback was slow, ambiguous, or entirely absent. Maybe, Kahneman and Klein began to suspect, they weren't actually disagreeing about whether intuition works. They were disagreeing about which environments make intuition trustworthy in the first place.¹⁶

That reframing became the spine of the 2009 paper. Kahneman and Klein proposed that the reliability of expert intuition depends on two conditions holding simultaneously: the environment has to have enough regularity and structure to be learnable in principle, what they called sufficient "validity," and the decision-maker has to have had a genuine opportunity to learn those regularities through prolonged practice with fast, clear feedback.²⁶ Where both conditions are met, expert intuition can be trusted almost the way you'd trust a well-calibrated instrument. Where either condition is missing, that same confident, fluent, effortless feeling of "just knowing" becomes actively dangerous, because subjective confidence, they stressed, is generated by the coherence of the story you can tell yourself, not by the accuracy of your judgment.⁷ A surgeon who has performed the same procedure a thousand times operates in what behavioral scientists call a "kind" environment — one where feedback is fast, reliable, and representative. A political analyst forecasting a coup, a fund manager picking individual stocks, a CEO betting the company on a market entry, these operate in "wicked" environments, ones with weak or delayed feedback and unstable rules, where fluent gut feeling is nearly worthless as a signal of accuracy.⁶⁷

This is the sense in which the paper truly was a failure to disagree. Kahneman didn't have to abandon thirty years of documenting bias. Klein didn't have to abandon his fireground commanders. They'd simply drawn a map with two territories on it, and each of them had spent their career exploring a different one.

What they still argued about, and why that matters more

It would be tidy, and slightly dishonest, to say the two men found perfect agreement. The McKinsey Quarterly interview conducted with both of them shortly after the paper's publication makes clear the disagreement didn't vanish, it just got more precise.⁷ Kahneman remained the more cautious voice throughout. Asked directly when executives can trust their gut, he answered that you should follow intuition under time pressure but otherwise treat it warily, because "overconfidence is a powerful source of illusions, primarily determined by the quality and coherence of the story that you can construct, not by its validity."⁷ Klein pushed back gently but firmly: intuition, he said, should never be blindly trusted, but it should be treated "as an important data point" worth deliberately evaluating rather than either obeying or dismissing outright.⁷

Their sharpest remaining disagreement was almost comically specific: checklists. Kahneman loved them, precisely because he distrusted intuition in low-validity environments, and a checklist doesn't require you to trust your gut, it substitutes procedure for judgment exactly where judgment is least reliable.⁷ Klein resisted them for the opposite reason: a checklist tells you the "then" of an if/then rule, but in a complex, ambiguous situation, the hard part is knowing whether the "if" condition has actually been satisfied, and that judgment call requires precisely the expertise a checklist can't encode.⁷ It's a small disagreement, but it's a perfect microcosm of the whole debate: Kahneman worried about the failure mode of false confidence, Klein worried about the failure mode of throwing away hard-won skill.

They agreed completely, however, on something that turned out to be more consequential than either checklists or intuition: the premortem. Klein's technique, in which a team imagines a project has already failed and works backward to explain why, got an enthusiastic endorsement from Kahneman, who called it "a great idea" and reported that when he presented it at Davos, a chairman of a major corporation said it alone was worth the trip.⁷ It's a lovely detail, because the premortem is neither a pure H&B tool nor a pure NDM tool. It's a structured way of harvesting expert intuition (what could plausibly go wrong, drawing on experience) while deliberately overriding the social pressure toward overconfidence that both men agreed was leadership's most dangerous default setting.⁷

Where this shows up outside the seminar room

The abstract framework, kind environments versus wicked ones, becomes much more vivid once you look at what happens when a carefully made plan meets a battlefield, a burning building, or a hospital at 3 a.m.

Consider the structure of a military campaign. Commanders spend enormous effort producing detailed plans of attack, precisely the kind of deliberate, analytical process the Heuristics and Biases tradition would recognize as good decision hygiene: gather intelligence, weigh courses of action, decorrelate sources of error, avoid groupthink. But the oldest maxim in military thinking, often attributed to Prussian field marshal Helmuth von Moltke, is that no plan survives first contact with the enemy. What happens in the minutes and hours after that contact, when smoke, casualties, and broken communications make the original plan partially or wholly obsolete, is not a failure of planning. It's the moment where Klein's entire research program becomes the operative one. The soldier or junior officer on the ground doesn't have time to build a decision matrix. They recognize the situation as a variant of something trained-for, mentally simulate whether the standard response will work here, and adapt it on the fly, exactly the recognition-primed pattern Klein documented in fireground commanders making split-second calls with lives on the line.⁵ The plan and the improvisation aren't opposed to each other. The plan is what stocks the pattern library that makes the improvisation possible in the first place.

Klein's own fireground research supplies an almost literal version of this. Studying incident commanders in high-stakes, fast-moving fires, Klein and his colleagues found that the more experienced commanders essentially never ran a formal comparison of options, they recognized the fire's behavior as a member of a familiar category, and that recognition triggered an appropriate course of action almost instantly, refined through a quick mental simulation rather than deliberate weighing of alternatives.³⁴ That is naturalistic decision-making in its purest form: not the absence of thought, but a compression of thought made possible by thousands of hours of prior experience with rich, fast feedback, exactly the "kind learning environment" Kahneman and Klein later agreed was the necessary condition for trustworthy intuition.⁶

It's worth noting where this same logic breaks down, because the breakdown is just as instructive. Kahneman's caution about "wicked" environments, ones with unstable rules and poor feedback, describes almost exactly the failure of intelligence forecasting in complex geopolitical crises, where analysts with decades of experience have repeatedly proven no better than chance at predicting specific political outcomes, precisely because the feedback loop that builds real expertise is too slow, too noisy, or too easily corrupted by hindsight. The NDM story explains why the fireground commander should be trusted. The H&B story explains why the geopolitical forecaster, despite feeling every bit as confident, should not be. Both stories are correct. They're just describing different territories on the map Kahneman and Klein eventually drew together.

Why H&B got the fame and NDM got the field

It's worth asking honestly why one of these schools became a household name and the other did not, because the answer says something uncomfortable about how ideas actually spread. Kahneman and Tversky's work traveled through cognitive illusions, small, self-contained puzzles that any reader could try on themselves and fail, producing the same delightful shock of recognition that made Thinking, Fast and Slow a bestseller. The bat-and-ball problem fits in a tweet. The recognition-primed decision model does not.

Klein's findings, by contrast, live in domains that resist compression: hours of fireground interviews, the tacit judgment of an ICU nurse who "just knows" a baby is crashing before the monitors show it, the pattern recognition of a chess grandmaster who sees the board differently than a novice does. NDM's evidence is thick, contextual, and hard to reduce to a clean laboratory demonstration, which makes it a harder story to tell at a dinner party but arguably a truer account of how most consequential decisions in the real world actually get made. Soldiers, surgeons, pilots, and emergency dispatchers don't spend their days solving urn-and-ball probability problems. They spend their days recognizing patterns under time pressure and improvising when the pattern breaks, which is precisely the terrain NDM was built to study.³⁴ If H&B explains why we should distrust our gut in the boardroom and the voting booth, NDM explains why we're often right to trust it in the emergency room and the fireground, and most people's working lives contain a lot more of the latter than the former.

What each school is still missing, and where the field is headed

Kahneman and Klein's détente didn't close the debate so much as sharpen the remaining open questions, and both traditions still have real gaps.

The Heuristics and Biases program is exceptionally good at demonstrating that a judgment deviates from a statistical norm, but weaker at telling you what to do about it in domains where no clean statistical norm exists, most real strategic decisions don't have a "correct" Bayesian answer sitting in the back of the textbook. Naturalistic Decision Making is exceptionally good at describing how genuine expertise functions once it exists, but historically weaker at explaining how to build that expertise faster, or at warning practitioners when they've wandered from a "kind" environment they trained in into a "wicked" one that merely resembles it, the exact trap that produces confidently wrong veteran forecasters and burned-out experts making calls outside their true competence.⁶⁷

The most promising work happening now sits explicitly in the overlap the 2009 paper mapped out. Klein has continued refining the recognition-primed decision model for domains like emergency medicine and military command, while researchers building AI-assisted decision tools have started asking Kahneman and Klein's diagnostic question, in effect, before deploying a system: is this a kind environment where pattern-matching intuition, human or algorithmic, deserves real trust, or a wicked one where structured checklists and decorrelated, independent judgments are the only defense against confident error.⁷

The likeliest future for this field isn't a victory for either camp. It's the slow spread of the Kahneman-Klein diagnostic itself, treating "should I trust this intuition" not as a philosophical question but as an empirical, checkable one: how predictable is this environment, and how good has the feedback been that trained the person, or the model, making the call? Klein's premortem and Kahneman's decorrelated independent judgments already coexist inside the same well-run organizations, not because someone declared a winner, but because each tool solves a different failure mode of the same underlying problem.⁷

The takeaway

Kahneman and Tversky earned the fame, and deservedly so, they gave the world an entire vocabulary for naming its own mistakes. But Gigerenzer, Klein, and the NDM tradition captured something the more famous story tends to leave out: that most of the decisions shaping human life, on firegrounds, battlefields, operating tables, and factory floors, are made not by weighing options against a formal model but by recognizing a situation and adapting, often correctly, in a fraction of a second. Neither school is the complete theory of how humans should decide under uncertainty. Heuristics and Biases is the necessary corrective for wicked, low-feedback environments where confidence regularly outruns competence, boardrooms, forecasting, and financial markets among them. Naturalistic Decision Making is the necessary corrective for the far more common situation the boardroom story tends to obscure, the kind, high-feedback environments where real expertise exists and overriding it with slow deliberation can be its own kind of error. The two traditions don't compete for the same territory so much as divide it between them, and the most useful question any decision-maker can ask isn't "should I trust my gut," but "what kind of environment am I actually standing in right now." That question, more than any single finding from either camp, is the real legacy of a paper that set out to disagree and, to its authors' evident surprise, couldn't quite manage it.

---

Footnotes

1. Conditions for intuitive expertise: a failure to disagree, PubMed. ↩ ↩²

2. Conditions for intuitive expertise: a failure to disagree, Semantic Scholar summary. ↩ ↩²

3. The Recognition-Primed Decision (RPD) Model, garyklein.com. ↩ ↩² ↩³

4. Rapid Decision Making on the Fire Ground, Klein, Calderwood & Clinton-Cirocco. ↩ ↩² ↩³ ↩⁴

5. Naturalistic Decision Making, Gary Klein, Human Factors, 2008. ↩ ↩²

6. Conditions for expert intuition, Merlin Rebrovic notes. ↩ ↩² ↩³ ↩⁴ ↩⁵

7. Strategic decisions: When can you trust your gut?, McKinsey Quarterly. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹²