Somewhere right now, a person who watched a four-minute video about immunology is absolutely destroying a doctor in a comment thread. They have fifteen upvotes. The doctor, exhausted and deeply aware of how much she doesn't know, has three.
This is not a bug in the internet's design. It is, somewhat horrifyingly, a predictable outcome of a documented cognitive phenomenon that existed long before WiFi: the Dunning-Kruger effect.
The short version: people with low knowledge in a field tend to dramatically overestimate how much they know. The internet, with its complete absence of barriers to publication, its engagement-rewarding algorithms, and its finely-tuned echo chambers, has turned this into something of a civilisational sport. Let's get into it.
What Is the Dunning-Kruger Effect, Actually?
In 1999, psychologists David Dunning and Justin Kruger at Cornell University published a paper with what might be the most politely devastating title in academic history: "Unskilled and Unaware of It: How Difficulties in Recognizing One's Own Incompetence Lead to Inflated Self-Assessments."
Their finding was elegant and a little bit grim. When they tested participants on logical reasoning, grammar, and the ability to spot what was funny (yes, they tested humour — academics have fun too), they found a consistent pattern: people who scored in the bottom quartile consistently estimated their own performance to be far above average. The worst performers were the most confident. The best performers, meanwhile, tended to slightly underestimate how well they'd done.
Why? Because the same skills that allow you to correctly assess a field are the skills you develop by learning that field. If you don't know much about, say, constitutional law, you also don't know how much constitutional law there is to know. You can't see the ocean from a puddle. The ignorance is, in a sense, self-concealing.
The Dunning-Kruger effect involves a double competence failure: (1) the person produces incorrect knowledge or conclusions, and (2) they lack the metacognitive ability to realise they've done so. It's not just wrong — it's confidently, cheerfully wrong. This is distinct from ordinary error, where the person knows something went sideways.
The Infamous Curve (And What It Really Means)
You've probably seen a version of the Dunning-Kruger curve. It usually looks something like this: confidence shoots up steeply with a tiny bit of knowledge, reaches a peak sometimes called "Mount Stupid" (which is extremely fair), then plummets into what researchers call the Valley of Despair as real understanding begins to accumulate and the person starts to grasp how much they don't know. Gradually, with sustained expertise, confidence grows back — but never quite reaches the giddy heights of knowing absolutely nothing.
The Stages of Knowing Things (Or Not)
The curve maps to recognisable stages that most people will have lived through in at least one domain — if they're honest with themselves, which, given the topic, is a big if.
Why the Internet Makes It 10× Worse
Here's where it gets interesting — or terrifying, depending on your disposition toward humanity.
The Dunning-Kruger effect existed long before the internet. People have been confidently wrong about things since at least the invention of writing, and probably much longer. What the internet did was remove virtually every structural mechanism that previously kept overconfident ignorance somewhat contained.
Before mass internet access, publishing something required either institutional backing (a publisher, a newspaper, a journal), access to broadcast infrastructure, or at minimum a printing press and distribution network. These were not always gatekeepers of quality, but they were gatekeepers of something — cost, access, friction.
The internet removed all of that friction. Which is great for access to knowledge and absolutely catastrophic for signal-to-noise ratio.
Every person on the internet publishes with the same visual authority as an institution that has existed for a century.
The Structural Problem with Open PublishingA tweet, a peer-reviewed paper, and a blog post by someone who read half a Wikipedia article all appear in essentially the same format, at the same size, with the same interface. There are no physical signals of credibility — no lab coat, no letterhead, no location in a university's institutional context. The design of the web, by default, equalises authority that is not equal.
The Five Internet Amplifiers
The Dunning-Kruger effect doesn't just survive on the internet — it thrives there because of five specific features of how online platforms are designed and how human psychology responds to them.
The result of these five combined is something new in human history: an information environment where the confidence with which something is stated has no reliable correlation to how accurate it is, and where the mechanisms that might correct this are weakened or absent.
Are You Doing This? (Spoiler: Probably)
This is the uncomfortable section. The Dunning-Kruger effect would be tidy if it only applied to other people. But that's not how it works — and research suggests the pattern is nearly universal in humans across at least some domains.
The good news is that it's domain-specific. You can be well-calibrated about your own professional area while simultaneously being on Mount Stupid about economics, nutrition, foreign policy, and whatever you got very confident about after reading that one long thread. High competence in one area does not transfer immunity.
🎯 Where Are You Right Now?
Pick a topic you have strong opinions about online. Move the slider to reflect how confident you feel. Then honestly ask yourself: how many primary sources have you actually read?
Reading about the Dunning-Kruger effect and thinking "yes, I understand this, it applies to other people" is itself a potential Dunning-Kruger moment. The people most immune to it are, almost by definition, those who hold their self-assessments most lightly. If you're nodding confidently at this article, that's worth sitting with for a second.
The Social Media Test
Here's a simple self-diagnostic. Think about the last time you wrote a confident comment, post, or reply online about a complex topic. Then ask:
• How many primary sources (not articles summarising primary sources) had you read?
• Could you explain the strongest argument against your position?
• Did your certainty level match your source depth?
For most people, most of the time, the honest answer to at least one of these is "no." That's not a personal failing — it's the default state of being human with internet access. The question is what you do with that recognition.
Online vs Offline Dunning-Kruger: A Comparison
| Factor | Offline | Online |
|---|---|---|
| Audience for overconfident claims | Limited — room, social group | Potentially global, permanently recorded |
| Speed of correction | Faster — social cues, expert presence | Slower — echo chambers delay feedback |
| Authority signals | Visible — credentials, context, institution | Stripped — all text looks equal |
| Incentive structure | Social cost for being confidently wrong | Engagement reward for confident claims |
| Exposure to dissenting views | Harder to avoid in daily life | Easy to algorithmically eliminate |
| Scale of influence | Contained to local social network | Viral potential regardless of accuracy |
Experts Are Also a Mess, Just Differently
In the interest of not creating a smug piece where the conclusion is "experts are right, laypeople are stupid," let's be clear about something important: the Dunning-Kruger effect doesn't exempt experts, it just deforms them differently.
As noted, people with high competence in one field often underestimate their ability — the opposite error. They've seen enough to know how much they don't know, which can make them artificially deferential. This is part of what the original research captured: the dual failure of metacognition goes in both directions.
More interestingly, experts are fully susceptible to Dunning-Kruger effects outside their domain. A brilliant engineer may be on Mount Stupid about economics. A distinguished economist may be overconfident about biology. Expertise is narrow; the internet asks everyone to comment on everything. The physicist commenting on geopolitics isn't using their physics knowledge — they're just a random person with opinions, like everyone else, but with a credential that grants false authority.
Later researchers raised methodological questions about the original study. Some argued the correlation between incompetence and overestimation was partly a statistical artefact of how the data was analysed. A 2020 paper by Gignac and Zajenkowski argued the effect, while real, is substantially smaller than the 1999 paper implied. The phenomenon exists — it's just more modest than the internet meme version. The irony of overestimating the size of an overconfidence study is not lost on anyone.
Can You Actually Fix This in Yourself?
Research has some things to say here, and they are simultaneously encouraging and annoying.
The encouraging part: metacognitive skills — the ability to accurately assess your own knowledge — are trainable. Studies show that people can become better calibrated over time, particularly through deliberate practice, feedback, and specific prompting to consider alternative views. Simply asking "what would a knowledgeable critic say about this?" before publishing an online opinion is documented to reduce overconfidence.
The annoying part: knowing about the Dunning-Kruger effect doesn't automatically make you immune to it. Awareness is a necessary but not sufficient condition. You can fully understand the research and still be on Mount Stupid about something an hour later. The bias operates below the level of abstract knowledge about it.
What does help, according to the evidence:
Read primary sources, not just summaries. The gap between "I understand this topic" and "I've read the actual research" is where most overconfidence lives. Steelman the opposition — before posting, try to articulate the strongest version of the counterargument. Track your predictions — if you make confident claims, note them and check back. Most people are far less accurate than they believe. Treat strong certainty as a warning flag, not a confirmation. Real expertise usually produces more uncertainty, not less.
- Kruger, J., & Dunning, D. (1999). Unskilled and unaware of it: How difficulties in recognizing one's own incompetence lead to inflated self-assessments. Journal of Personality and Social Psychology, 77(6), 1121–1134. doi.org
- Gignac, G. E., & Zajenkowski, M. (2020). The Dunning-Kruger effect is (mostly) a statistical artefact: Valid approaches to testing the hypothesis with individual differences data. Intelligence, 80, 101449. doi.org
- Vosoughi, S., Roy, D., & Aral, S. (2018). The spread of true and false news online. Science, 359(6380), 1146–1151. doi.org
- Dunning, D. (2011). The Dunning–Kruger effect: On being ignorant of one's own ignorance. Advances in Experimental Social Psychology, 44, 247–296. doi.org
- Pennycook, G., & Rand, D. G. (2019). Lazy, not biased: Susceptibility to partisan fake news is better explained by lack of reasoning than by motivated reasoning. Cognition, 188, 39–50. doi.org
- Fischhoff, B. (1977). Perceived informativeness of facts. Journal of Experimental Psychology: Human Perception and Performance, 3(2), 349–358. doi.org
- Pariser, E. (2011). The Filter Bubble: What the Internet Is Hiding from You. Penguin Press. ISBN: 978-1594203008.
- Dunning, D. (2014). We Are All Confident Idiots. Pacific Standard Magazine. psmag.com
This article is intended for informational and educational purposes only. The psychological research referenced is drawn from publicly available peer-reviewed academic sources as cited above. The Dunning-Kruger effect is a documented area of psychological research, though the strength and universality of the effect continues to be discussed and refined in the academic literature. No individuals, organisations, or groups are identified or targeted. This content does not constitute professional psychological, medical, or any other professional advice. Readers are encouraged to consult the original cited sources for full academic context. Views presented are for educational discussion and general understanding.