The Dunning–Kruger Effect Is Not a Meme. It Is a Warning About Who Gets to Talk With Certainty

The popular version of the Dunning–Kruger effect is a punchline: stupid people are loud, smart people are humble. That is not what the research says, and it is a sloppy way to run a public argument. The actual claim is narrower and more useful. In domains that require skill, the same gaps that produce bad performance also interfere with a person’s ability to see that the performance is bad.¹

That is a metacognitive problem, not a moral one. It matters in laboratories. It matters more in courtrooms, comment files, and policy fights, where confidence is often treated as a substitute for a record.

What Kruger and Dunning Actually Found

In 1999, Justin Kruger and David Dunning reported four studies on humor, grammar, and logic. Participants in the bottom quartile scored around the 12th percentile and estimated themselves near the 62nd.² The authors’ explanation was a double burden. Low skill produces errors. The same low skill removes the tools needed to detect those errors.³

They were not saying that every confident person is incompetent. They were saying that competence and self-assessment use overlapping machinery. If you cannot tell a valid inference from an invalid one, you also cannot tell whether your own inference is valid.⁴ Later work in the same line found that training the underlying skill improved calibration. People did not become modest by being scolded. They became more accurate after they acquired the standard they had been missing.⁵

Dunning later described the broader pattern as meta-ignorance: ignorance of the scope of one’s own ignorance. Gaps hide among “unknown unknowns,” or they are covered by beliefs that feel like knowledge because they are familiar.⁶ Overconfidence in that setting is not theatrical. It is what incomplete maps feel like from the inside.

What the Effect Is Not

Three popular uses should be dropped.

It is not a diagnosis you get to hang on an opponent. The original studies measured test performance against self-ratings in defined tasks. They did not license armchair clinical judgments about a person’s character.⁷

It is not proof that experts are always humble. High performers in the original work were more calibrated than the bottom quartile, but they sometimes underestimated their relative standing.⁸ Expertise reduces one kind of error. It does not abolish self-interest, tribal loyalty, or the urge to win a room.

It is not settled as a giant, universal curve. A serious literature argues that part of the famous graph is a statistical artifact: regression to the mean plus a general better-than-average bias can produce a similar picture even without a special metacognitive failure at the bottom.⁹ Gignac and Zajenkowski applied stricter tests and concluded that the classic effect is smaller than the meme, and in some datasets mostly artifactual.¹⁰ Other teams, using different samples, still find a reduced but real effect.¹¹ The honest summary is not “debunked” and not “law of nature.” It is: poor performers often misjudge themselves; the size and cause of that misjudgment depend on the task and the method.

That caution is the point. If you use Dunning–Kruger as a club, you are doing the thing the paper warned against: treating a complicated finding as if fluency were the same as command.

Why This Shows Up in Policy Fights

Public advocacy rewards the same cues the lab treats as unreliable. Volume. Certainty. A story that fits the tribe. A claim that “nobody else understands the law.” Those cues are cheap. The skills that would audit them — reading a statute against a regulation, matching a citation to a proposition, distinguishing an existing rule from a proposed amendment — are expensive.¹²

So the pattern repeats. Someone who has lived inside a program for years can still be unskilled at the legal question the year actually poses. Longevity is not metacognition. A person can be fluent on a Facebook page and lost in SDCL 1-26-4.11. A person can be sincere about patients and still unable to tell a finding of fact from a press line. The Dunning–Kruger literature predicts that the person will not experience that gap as a gap. It will feel like expertise.¹³

Courts and agencies are built, at least on paper, to interrupt that feeling. They ask for a record. They ask whether the evidence supports the element. They ask whether the comment addresses the notice. They do not ask who sounded the most certain in the hallway. When those systems work, they are a mechanical answer to meta-ignorance: you do not have to trust the speaker’s self-score. You can look at the paper.¹⁴

When those systems are used as theater — filings as punishment, hearings as content, rulemaking as branding — the bias comes back. The room starts scoring performance again. That is when overconfident error becomes expensive for other people, because the cost lands on patients, licensees, and the next person who has to live under the rule.

The Better Test Than the Meme

If the research is used honestly, it supplies a test, not an insult.

Ask what skill is actually required. Grammar is not the same as humor. Running a campaign is not the same as briefing an appellate standard. Owning a shop is not the same as reading a regulatory impact analysis. Collapse those categories and everyone looks like an expert in everything.¹⁵

Ask what would change the speaker’s mind. Kruger and Dunning’s most useful result was that adding skill improved insight.¹⁶ A person who cannot name the document that would falsify the claim is not in a dispute about values. They are in a dispute about whether a map exists.

Ask whether the claim is calibrated to the authority. “This packet should state whether it models 108 establishments or 139” is a bounded ask. “I already know what federal law means and the rest of you are lying” is an unbounded self-score. The first can be checked against a .gov page. The second cannot.¹⁷

Recent work in the same research family finds that actual knowledge reduces overclaiming — the habit of asserting familiarity with things that do not exist — while mere self-perceived expertise increases it.¹⁸ That split is the whole subject. Real command and the feeling of command are not the same inventory.

What WeedPress Is For, In This Light

A public file is a device for getting around other people’s self-assessments, including your own. Transcripts, establishment lists, Register notices, and comment letters do not care who felt informed. They either support the proposition or they do not.

That is why the useful application of Dunning–Kruger in South Dakota cannabis policy is not a personality essay. It is a method. Put the numbers in the packet. Cite the section that is already in force before you describe the amendment. Separate the statute that belongs in a different proceeding. Leave the self-score out of the ask.¹⁹

The research will keep arguing about how large the effect is. That argument is healthy. The operational lesson does not depend on winning it. Confidence is easy to manufacture. Calibration has to be earned against a standard the speaker does not control. In a rulemaking, that standard is the record. Treat anyone — including the person writing this — as untrustworthy until the paper is in the file.

Footnotes

¹ Justin Kruger & David Dunning, Unskilled and Unaware of It: How Difficulties in Recognizing One’s Own Incompetence Lead to Inflated Self-Assessments, 77 J. Personality & Soc. Psychol. 1121, 1121–22 (1999).

² Id. at 1123–28.

³ Id. at 1121–22.

⁴ David Dunning, Chip Heath & Jerry M. Suls, Flawed Self-Assessment: Implications for Health, Education, and the Workplace, 5 Psychol. Sci. in the Pub. Int. 69, 70–73 (2004).

⁵ Kruger & Dunning, supra note 1, at 1128–31; Joyce Ehrlinger, Kerri Johnson, Matthew Banner, David Dunning & Justin Kruger, Why the Unskilled Are Unaware: Further Explorations of (Absent) Self-Insight Among the Incompetent, 105 Organizational Behav. & Hum. Decision Processes 98 (2008).

⁶ David Dunning, The Dunning–Kruger Effect: On Being Ignorant of One’s Own Ignorance, 44 Advances in Experimental Soc. Psychol. 247, 247–50 (2011).

⁷ See Kruger & Dunning, supra note 1, at 1121–23 (describing experimental tasks, not clinical assessment).

⁸ Id. at 1123–27.

⁹ Joachim Krueger & Ross A. Mueller, Unskilled, Unaware, or Both? The Better-Than-Average Heuristic and Statistical Regression Predict Errors in Estimates of Own Performance, 82 J. Personality & Soc. Psychol. 180 (2002); Thomas Schlösser, David Dunning, Kerri L. Johnson & Justin Kruger, How Unaware Are the Unskilled? Empirical Tests of the “Signal Extraction” Counterexplanation for the Dunning–Kruger Effect, 39 J. Econ. Psychol. 85 (2013).

¹⁰ Gilles E. Gignac & Marcin Zajenkowski, The Dunning-Kruger Effect Is (Mostly) a Statistical Artefact: Valid Approaches to Testing the Hypothesis with Individual Differences Data, 80 Intelligence 101449 (2020).

¹¹ Curtis S. Dunkel, Joseph Nedelec & Dimitri van der Linden, Reevaluating the Dunning-Kruger Effect: A Response to and Replication of Gignac and Zajenkowski (2020), 96 Intelligence 101717 (2023).

¹² See generally Dunning, Heath & Suls, supra note 4, at 90–98 (workplace and professional self-assessment).

¹³ Dunning, supra note 6, at 250–60.

¹⁴ Cf. SDCL 1-26-4 (requiring notice, hearing, and a written-comment process for proposed permanent rules); SDCL 1-26-4.11 (regulatory impact analysis).

¹⁵ Kruger & Dunning, supra note 1, at 1122–23 (domain-specific tasks).

¹⁶ Id. at 1128–31.

¹⁷ See Dunning, supra note 6, at 247–49.

¹⁸ Stav Atir, Emily Rosenzweig & David Dunning, Does Expertise Protect Against Overclaiming False Knowledge?, 184 Organizational Behav. & Hum. Decision Processes 104354 (2024).

¹⁹ See Kruger & Dunning, supra note 1, at 1131–32; SDCL 1-26-4.