The Dunning–Kruger Misconception on Social Media and the Importance of Intellectual Humility

1999, Cornell University: “Unskilled and Unaware of It”

The Dunning–Kruger effect is based on a study published in 1999 by Justin Kruger and David Dunning of Cornell University in the Journal of Personality and Social Psychology . Even the article's title is clear enough to stick in your mind: “Unskilled and Unaware of It: How Difficulties in Recognizing One's Own Incompetence Lead to Inflated Self-Assessments.” The study can be summarized as follows: participants took tests of logic, grammar, and appreciation of humor; they then answered the question “Where do I rank on this test?” The result was striking. The worst performers (actually in the bottom %12) thought they were above average (at the %62 percentile). High performers, meanwhile, made a slight “humble error,” estimating their own ranking a little lower.

Dunning and Kruger explained this pattern with the “double burden” hypothesis: when knowledge in a field is limited, performance is not the only thing that suffers; the higher-order cognitive ability needed to recognize one's own errors (metacognition) is also lacking. In other words, not knowing also brings the problem of “not knowing that you do not know.” This is actually a pattern we have long recognized in clinical medicine — experienced clinicians commonly observe that the doctor who knows very little about a subject may make the most decisive clinical judgments about it.

Statistical Criticism: Does the Effect Really Exist?

One of the most heated debates in the scientific literature over the last five years concerns the mathematical robustness of the Dunning–Kruger effect itself. A study published in 2020 by Gilles Gignac and Marcin Zajenkowski in Intelligence , entitled “The Dunning-Kruger Effect Is (Mostly) a Statistical Artefact,” suggested that a large part of the effect could arise from two statistical mechanisms:

This criticism remains under discussion. Avram Hiller's 2023 commentary questioned Gignac & Zajenkowski's method; the 2023 replication study by Dunkel, Nedelec, and van der Linden concluded that “the effect exists, but it is weak.” Thus, as of 2026, the scientific community agrees on this: the Dunning–Kruger effect is not zero, but it is not as dramatic as the original 1999 graphs suggested. It is a smaller, more domain-specific observation, intertwined with multiple mechanisms.

Studies from 2021 and 2024: An Important but Unexaggerated Association

The biggest misconception in the news circulating in Turkish media is the claim that “Dunning–Kruger research introduced the concept of intellectual humility.” No — these are two separate research fields. The concept of intellectual humility has developed into a discipline in its own right over the past decade, and two studies in particular have tried to place it in the context of Dunning–Kruger:

So the actual science says this: Intellectual humility matters. It reduces overconfidence. It supports critical thinking. But no primary paper has called it “the most reliable sign of high intelligence”. That sentence is a dramatic transformation created by the news text shared on social media — it appears in none of the studies by Leman, Bowes, or Krumrei-Mancuso.

A Five-Point Comparison

There is something noteworthy here: when explaining the dangers of overconfidence, the intellectual-humility literature itself frequently refers to medical diagnostic errors . The landmark reference in this literature is the classic paper published in 2008 by Eta Berner and Mark Graber in The American Journal of Medicine.

Berner and Graber, 2008: Physician Overconfidence Is a Public Health Problem

The 2008 review by Eta Berner and Mark Graber is a pioneering study that systematically shows how physician overconfidence disrupts medical decision-making. It has three main findings:

Then, in 2013, Ashley Meyer and Hardeep Singh's team confirmed this pattern in a study published in JAMA Internal Medicine: between physicians' diagnostic accuracy and their confidence in their diagnoses, the correlation is weak. In other words, a doctor who appears more certain of a diagnosis is not necessarily more correct. Indeed, in difficult cases, accuracy does not remain unchanged alongside that certainty — it falls. Overconfidence does not increase diagnostic accuracy; it reduces the tendency to request additional tests, discourages seeking a second opinion, and prevents patients from benefiting from clinical decision-support tools.

The Words “I May Be Wrong” Can Save a Life

Oncology is one of the most complex fields of modern medicine. Even a diagnosis of lung cancer no longer represents a single disease; EGFR, ALK, ROS1, KRAS G12C, MET exon 14, BRAF V600E, NTRK, RET, HER2 define a molecular family with at least nine subgroups. Each has a different treatment algorithm. A doctor initiating treatment on the basis of “lung cancer = chemotherapy” was standard practice until 2010, but today it is a clinical error. In the era of precision oncology, treating without finding out the tumor's identitybelongs to a past that oncology has left behind.

In clinical practice, this humility appears at three different levels:

“I Cured Cancer in 3 Weeks” — The Most Dangerous Stage for Dunning–Kruger

One of the problems oncologists encounter most frequently in everyday clinical practice is claims circulating on social media such as “I cured cancer in 3 weeks with this plant” . These posts — unrelated to science, delaying patients' treatment, and sometimes resulting in death — are the Dunning–Kruger effect in its purest form. The people with the least knowledge of the subject speak with the greatest confidence. A physician who understands the true complexity of medical oncology could never make a statement such as “it definitely cures cancer.”

This is why an oncologist's job is not only to provide the right treatment , but also to support patient literacy. A consultation in which a patient arrives saying “Doctor, I read this on the internet” is an opportunity for clinical communication — and the physician must evaluate the patient's information sources without belittling or looking down on them, but by placing them on a scientific footing . This, too, is a form of humility: taking the patient's search for information seriously, offering an evidence-based framework, and, when necessary, being able to say “I do not know, but I will look into it.”

ChatGPT, Claude, and Medical Hallucinations

As of 2026, artificial-intelligence language models (ChatGPT, Anthropic Claude, Google Gemini, Med-PaLM) have become sources consulted by both physicians and patients. However, their best-known weakness is the phenomenon called hallucination : in a confident tone, the model can generate references, studies, dosages, or treatment protocols that do not actually exist. Essentially, this reflects the Dunning–Kruger effect in the context of artificial intelligence — a model not knowing its own limitations and not recognizing that what it produces is wrong.

Interestingly, this issue is equally important for everyone, from oncologists to patients. When a patient asks ChatGPT “Is this treatment suitable for me?”, critically reading the model's answer requires a form of intellectual humility. Similarly, turning an AI-generated literature summary into a clinical decision without verification would, for a physician, represent a failure of their own metacognition .

This article provides information about scientific literacy and clinical epistemology; it does not contain individual treatment recommendations. Editor's note: accessing and verifying the source papers behind headlines circulating on social media is a fundamental literacy skill for every reader in the modern digital age. In oncology, intellectual humility is not an individual virtue but an institutional indicator of clinical quality — multidisciplinary conferences, NGS-based decision-making processes, and the institutionalization of a second-opinion culture are concrete expressions of this humility.