---
id: "20260812-0230-what-does-kahneman-klein"
title: "What does Kahneman & Klein (2009) actually say about intuition failing in low-validity environments?"
type: "capture"
status: "promoted"
promoted_to: ["30-notes/claim-kahneman-klein-2009-skill-cannot-form-in-low-validity-confidence-does-not-signal-validity.md","30-notes/claim-kahneman-klein-2009-name-stock-and-political-forecasting-as-zero-validity-environments.md","30-notes/claim-kahneman-klein-2009-confidence-tracks-internal-consistency-not-evidence-quality.md","30-notes/claim-kahneman-klein-2009-intuition-reliable-only-in-high-validity-environments.md","50-questions/question-verify-kahneman-klein-2009-low-validity-intuition-quote-primary.md"]
not_promoted: ["Fractionated expertise (nurses/physicians/auditors skilled in part of a domain, not all) — filed by the capture itself under 'Further leads', a supporting detail, not a distinct load-bearing claim this session. Left as a lead.","Grove, Zald, Lebow, Snitz & Nelson (2000) meta-analysis (mechanical > clinical judgment in ~half of 136 studies) — cited SECONDHAND by Kahneman & Klein and carrying [unverified-quant — needs primary]. A quantitative claim may not rest on a secondhand citation; needs the Grove et al. primary. Not routed to 50-questions/ (not load-bearing for any kept claim — no promise made).","Klein's 'premortem' method (Klein 2007, HBR) — a lead for a separate debiasing claim-note, not in this paper's evidentiary core. Left as a lead.","'Automation bias' (Skitka, Mosier & Burdick 1999/2000) — cited lead on algorithm-oversight failure; separate source, not promoted here.","USS Vincennes (1988) shootdown as founding catalyst of the NDM/TADMUS program — a historical lead the capture did not pursue; not promoted.","Entity candidates (Paul Meehl, Adriaan de Groot, Egon Brunswik, Robin Hogarth, Daniel Kahneman, Gary Klein, illusion of validity, zero-validity environment, NDM/HB) — all first-appearance in the vault this session; per the entity-page spec's bias-against-the-flood, none promoted to hubs from a single capture. Strongest structural candidate (Meehl) flagged [entity] in seek-flags.md. entity-herbert-simon already covers the Simon build-on; left untouched (see journal)."]
origin: "batch"
writer_model: "claude-sonnet-5"
date_created: "2026-08-12T00:00:00.000Z"
provenance: "batch run, 2026-08-12"
derived_from: []
tags: ["kahneman","klein","expert-intuition","low-validity-environment","judgment-under-uncertainty","cognitive-psychology","illusion-of-validity","verification"]
source_url: "https://gwern.net/doc/psychology/cognitive-bias/illusion-of-depth/2009-kahneman.pdf"
source_sha: "4ca2448fc314353d8af78a495208f6e9b009e03e5dc425fffc02c7c3cdda1ba5"
source_author: "Daniel Kahneman and Gary Klein"
source_date: "2009-09"
source_title: "Conditions for Intuitive Expertise: A Failure to Disagree"
source_venue: "American Psychologist, Vol. 64, No. 6, 515–526 (mirror hosted by gwern.net; the APA/PsycNET original at doi.org/10.1037/a0016755 is paywalled and no author-hosted copy on kahneman's or klein's own domain was located)"
source_tier: 1
seek_code_commit: "729ee25"
---


This capture directly answers [[question-verify-kahneman-klein-2009-low-validity-intuition-quote-primary]], the open verification question raised against [[claim-kahneman-klein-2009-intuition-reliable-only-in-high-validity-environments]]. That existing claim-note was sourced only through a bibliographic reference-index page (scirp.org) and paraphrased the paper's conclusion as "confident intuition becomes systematic error" in low-validity environments — a phrase that does not appear in the primary text. The paper was located, fetched via `extract_pdf` (a gwern.net-hosted mirror of the American Psychologist PDF; tls verified, sha256 recorded above), and read directly. **The central question is answered**: the paper does discuss low-validity environments and intuition failure, but its own framing is narrower and more precise than "systematic error" — the paper's claim is that skilled intuition cannot *develop* in low-validity conditions, and that subjective confidence gives no reliable signal for telling a skilled intuition from an unskilled one, in any environment.

## Claim: Skilled intuition requires two necessary conditions — an environment of sufficiently high validity, and adequate opportunity to learn its regularities through prolonged practice and rapid, unequivocal feedback

**Claim type**: definitional / technical-mechanism. **Source tier required**: Tier 1–2 (technical mechanism) — met, Tier 1, primary document, direct quote.

The paper states this as its central positive finding, in the Conclusions section: "An environment of high validity is a necessary condition for the development of skilled intuitions. Other necessary conditions include adequate opportunities for learning the environment (prolonged practice and feedback that is both rapid and unequivocal). If an environment provides valid cues and good feedback, skill and expert intuition will eventually develop in individuals of sufficient talent." Earlier in the body: "Two conditions must be satisfied for skilled intuition to develop: an environment of sufficiently high validity and adequate opportunity to practice the skill." "Validity," as the authors define it, "describes the causal and statistical structure of the relevant environment" — i.e., whether stable, learnable relationships exist between observable cues and outcomes.

This is the primary-source confirmation of the scope condition that [[claim-simon-defines-expert-intuition-as-domain-bound-recognition|Simon's "intuition is recognition"]] claim rests on: the paper explicitly builds on Simon (1992), quoting his definition directly ("The situation has provided a cue: This cue has given the expert access to information stored in memory, and the information provides the answer. Intuition is nothing more and nothing less than recognition") and adds the two boundary conditions above as the test for when that recognition process is trustworthy.

## Claim: In low- and zero-validity environments, true skill cannot develop — but people still produce confident intuitions there, and subjective confidence does not distinguish valid intuitions from invalid ones, in any environment

**Claim type**: technical-mechanism. **Source tier required**: Tier 1–2 — met, Tier 1, direct quote.

The paper is explicit that the failure mode is not that intuition becomes actively worse or more "systematic" in error under low validity — it is that skill cannot form there at all, while the subjective experience of confident intuition persists regardless: "Although true skill cannot develop in irregular or unpredictable environments, individuals will sometimes make judgments and decisions that are successful by chance. These 'lucky' individuals will be susceptible to an illusion of skill and to overconfidence." And, as a general conclusion not restricted to low-validity cases: "Subjective confidence is therefore an unreliable indication of the validity of intuitive judgments and decisions." Elsewhere: "high subjective confidence is not a good indication of validity," and "we do not believe that subjective confidence reliably indicates whether intuitive judgments or decisions are valid... people do not have a strong ability to distinguish correct intuitions from faulty ones."

This refines (rather than confirms) the existing vault claim-note's paraphrase. "Systematic error" is not the paper's own language; the paper's claim is more specific — no skill forms, confidence is uninformative, and correct-by-chance judgments are indistinguishable, subjectively, from earned ones.

## Claim: The paper names specific domains as (near-)zero-validity — individual stock-price prediction and long-term political forecasting — where outcomes are effectively unpredictable and confident intuition cannot be earned skill

**Claim type**: historical / definitional (the authors' own worked examples). **Source tier required**: Tier 3–4 acceptable for definitional/example claims; sourced here at Tier 1 anyway.

"[O]utcomes are effectively unpredictable in zero-validity environments. To a good approximation, predictions of the future value of individual stocks and long-term forecasts of political events are made in a zero-validity environment." The paper contrasts this with high-validity domains it names explicitly as supporting genuine skill: "Medicine and firefighting are practiced in environments of fairly high validity." It also cites Tetlock's (2005) finding that experienced political forecasters were not superior to untrained newspaper readers at long-range political forecasting, attributing the failure to the environment rather than to the forecasters: "the problem is in the environment: Long-term forecasting must fail because large-scale historical developments are too complex to be forecast."

## Claim: The mechanism behind unwarranted confidence is that subjective confidence tracks the internal consistency of the evidence a judgment is based on, not the quality or validity of that evidence — so redundant, flimsy evidence produces overconfident judgments

**Claim type**: technical-mechanism. **Source tier required**: Tier 1–2 — met, Tier 1, direct quote.

"Subjective confidence is often determined by the internal consistency of the information on which a judgment is based, rather than by the quality of that information... As a result, evidence that is both redundant and flimsy tends to produce judgments that are held with too much confidence. These judgments will be presented too assertively to others and are likely to be believed more than they deserve to be." This is the paper's proposed causal account of *why* confidence fails as a validity signal specifically in low-validity settings: low-validity environments are exactly where the available cues are weak, so any apparent internal consistency among them is spurious rather than earned, yet it still produces high subjective confidence.

## Further leads

- The paper explicitly names "fractionated expertise" — professionals (their examples: nurses, physicians, auditors) who have genuine skill in some parts of their domain but not others, a source of overconfidence when skilled judgment is misapplied outside its validated zone. Same source as above, pp. 522–523.
- Grove, Zald, Lebow, Snitz & Nelson (2000), a meta-analysis of 136 studies, is cited by Kahneman & Klein for the quantitative claim that mechanical/algorithmic judgment outperformed clinical judgment in about half of studies and tied in most of the rest — cited secondhand here, would need the Grove et al. primary before any quantitative claim-note. [unverified-quant — needs primary]
- Klein's "premortem" method (Klein, 2007, *Harvard Business Review*) is presented as a debiasing technique both authors endorse — a lead for a separate claim-note on debiasing organizational decisions.
- "Automation bias" (Skitka, Mosier & Burdick, 1999/2000) — cited by Kahneman & Klein as the risk that human supervisors of algorithms become passive/less vigilant; a lead on algorithm-oversight failure modes.
- The paper's own account of the USS Vincennes 1988 shootdown as the founding catalytic event for the naturalistic-decision-making research program (TADMUS) — a historical lead, not pursued here.

## Entity candidates

- Paul Meehl — person — the foundational figure the heuristics-and-biases (HB) side of this paper's comparison is built against: Kahneman & Klein trace the entire HB skeptical-of-expertise tradition to Meehl's 1954 monograph comparing clinical vs. statistical prediction, and this paper's Grove et al. (2000) discussion is a direct descendant of the "Meehl paradigm." Older and more load-bearing to this paper's ancestry claim than either living co-author.
- Adriaan de Groot — person — the foundational figure the naturalistic-decision-making (NDM) side is built against: his 1946/1978 studies of chess grandmasters are the paper's cited origin point for "intuition as recognition," predating and underlying both Simon's and Chase & Simon's later formalizations.
- Egon Brunswik — person — credited by name as the originator of the environmental-validity framework ("the ways in which skilled judgments take advantage of environmental regularities have been discussed by, among others, Brunswik, 1957") that the paper's central "validity" concept is built on.
- Robin Hogarth — person — coined "wicked environments" (vs. "kind" environments), the term the paper uses for domains where learned regularities are actively misleading; cited repeatedly as a direct precursor to this paper's own conclusions.
- Daniel Kahneman — person — co-author; already extensively covered in this vault's [[claim-kahneman-klein-2009-intuition-reliable-only-in-high-validity-environments]] cluster.
- Gary Klein — person — co-author; originator of the recognition-primed decision (RPD) model discussed throughout.
- illusion of validity — concept — term Kahneman coined (from his own IDF officer-selection experience, described in this paper) for unjustified confidence in judgment that persists despite disconfirming statistical feedback.
- zero-validity environment — concept — the paper's own term for domains (stock-price prediction, long-range political forecasting) where no learnable structure exists at all.
- naturalistic decision making (NDM) / heuristics and biases (HB) — concept — the paper's frame for the two research traditions it attempts to reconcile.