EN
The cover of Sources of Power on a wooden table beside geometric pieces and a magnifying glass.

Sources of Power: when expert intuition deserves a place in assessment

Sources of Power makes it possible to treat expert intuition as situated recognition without turning experience or confidence into unquestionable authority.

PRYSMAP6 min read

The publisher’s description of Gary Klein’s book places his work in decisions made under pressure by experienced people. This article is not based on the book’s full text. The applied interpretation is limited to that public thesis and the later article Klein wrote with Daniel Kahneman.

That limit matters because “expert intuition” is often used to justify two opposing positions. One dismisses it as bias. The other demands obedience because it comes from years of experience. Klein’s work opens a third path: study what the person recognizes, where they learned the pattern, and how they test it.

Experience compresses a history of signals

An experienced professional does not observe an empty situation. She perceives sequences, absences, and combinations that another person cannot yet distinguish.

In a complex assessment, that recognition may explain why a technically correct decision was fragile. Perhaps the person reached the outcome but ignored an early signal or needed someone else to correct the path.

A general rubric may lose that distinction. Excluding expert judgment impoverishes the interpretation. Accepting it without examination creates a different problem: no one can know whether the conclusion comes from a valid pattern, reputation, or personal preference.

The environment determines what intuition can learn

Kahneman and Klein compared the heuristics-and-biases tradition with naturalistic decision-making. Their central agreement sets conditions for expert intuition: an environment with sufficiently valid regularities and opportunities to learn through practice and feedback. Kahneman and Klein, Conditions for Intuitive Expertise: A Failure to Disagree, 2009.

The authors’ agreement shifts the question away from subjective confidence and toward the structure of learning. Feeling certain does not show that the environment offered valid patterns or timely correction.

This requires asking about the domain. Someone may have sound intuition for frequent technical incidents and little basis for predicting another person’s performance in a new role.

Years of experience do not settle the question either. What matters is the variety of cases, the quality of feedback, and the opportunity to correct errors.

Feedback must also arrive with enough resolution. Knowing that an intervention ended well does not explain which signal was useful or which alternative would have failed. Intuition learns more effectively when an outcome can be linked to specific decisions, not when it receives general approval months later.

Two firefighters inspect the blackened interior of a structure and point to marks on the wall.

This creates boundaries within the same profession. The regularity that makes it possible to anticipate a known failure may not exist when judging an innovation, a new market, or a role transition. Expertise needs to state its territory before extending into another decision.

The MIT Press page presents Sources of Power as a study of real-world decisions by experienced people. It highlights experience and recognition. That territory organizes this interpretation, but it does not authorize attributing details from a book that was not consulted. Gary Klein, Sources of Power, MIT Press.

Externalizing judgment preserves its value

A specialist says that someone is not yet ready for a responsibility. Instead of asking her to replace the impression with a formula, the assessment can request four elements: the signal observed, the expected option, the anticipated risk, and comparable cases.

The explanation does not need to exhaust the reasoning. It must offer a basis that another person can examine and the person being assessed can understand.

If the judgment changes when the episodes are reconstructed, the process produced learning. If it holds, there is now a path from evidence to conclusion.

Externalizing also makes it possible to compare experts. Two people may recognize different signals because they learned in different contexts. The discrepancy helps delimit the domain and determine what additional evidence would resolve the interpretation.

It also reveals when an expert is applying a rule learned under conditions that have since changed. If the signals, tools, or consequences are new, the explanation provides a point for review. Without it, the organization discovers the mismatch only after the outcome.

Familiarity and expertise follow different paths

Working with someone for many years increases access to episodes. It also creates exposure to halo effects, reputation, and selective memory.

A statement such as “I know them” does not show that the source observed the relevant manifestation. The authority to assess comes from the role; the strength of the judgment depends on relevant evidence.

This limit connects the review with Self-assessment, manager, or peers—what each source contributes. Every source needs access to the object being assessed and rules appropriate to the later use.

The structure needs experts too

A shared architecture establishes what is assessed, which differences matter, and which uses are authorized. The expert interprets complex signals within that territory.

The combination avoids two extremes. A rigid scale may ignore patterns it did not anticipate. Unconstrained judgment may turn status into evidence.

In sensitive decisions, another expert should be able to review the chain. Insufficient-evidence states must remain available, and the conclusion needs a route for challenge.

The review is not meant to produce a vote among specialists. It should verify access to episodes, relevance of the domain, and the connection between signals and conclusion. If disagreement persists, the response may be to limit the use of the judgment or produce additional evidence, not to manufacture consensus.

The organization can record patterns without creating permanent surveillance. It needs a purpose, limited access, and proportionate retention. The goal is to make judgment more defensible, not to accumulate impressions about every behavior.

Expert intuition deserves a place when it improves the quality of hypotheses and makes its signals open to examination. To keep that place, assessment must require signals, limits, and a route for challenge; an inability to explain cannot serve as proof of authority.

To extend this reading, see How to design a calibration session that does not turn into a negotiation over scores, What it means for an assessment to be valid for a specific decision, and Confidence is not competence, which develop complementary dimensions of the problem.