Dunning-Kruger in the Group Chat: Why the Loudest Friend Is Rarely the Wisest

The Dunning-Kruger effect is a cognitive bias, described by psychologists Justin Kruger and David Dunning in 1999, in which people with limited skill in a specific area overestimate their ability, because the knowledge needed to perform well is the same knowledge needed to recognize good performance. In conversation, it shows up as confident takes on topics someone barely knows.

The study has a famously strange origin story. In 1995, a man in Pittsburgh robbed two banks in broad daylight without a disguise. When police showed him the security footage, he was stunned. He had rubbed lemon juice on his face, convinced it would make him invisible to cameras, the same way lemon juice works as invisible ink. He had even "tested" it with a Polaroid first.

Dunning read about the case and asked a sharper question: what if being bad at something also makes you bad at *knowing* you're bad at it?

He and Kruger, then a graduate student at Cornell, tested the idea in a paper titled *"Unskilled and Unaware of It: How Difficulties in Recognizing One's Own Incompetence Lead to Inflated Self-Assessments,"* published in the *Journal of Personality and Social Psychology*. Students took tests of humor, logical reasoning, and English grammar, then guessed how they had done compared with their peers.

The pattern was striking. Students who scored in the bottom quarter consistently placed themselves comfortably above average.

Now translate that to your group chat. Nobody is robbing banks with citrus. But somebody is explaining how the Federal Reserve works after one podcast, rating their own karaoke a solid 9, or insisting they "basically know" how to get to the restaurant. The confidence isn't fake. It's sincere. That's what makes it the Dunning-Kruger effect instead of just bluffing.

A metacognitive deficit produces unwarranted confidence because judging an answer requires the same skill as producing it. Dunning calls this the dual burden: when you lack expertise, you both make more mistakes and lose the ability to spot them. Without that internal error detector, your work feels fine, so your confidence never gets the signal to drop.

**Metacognition** is thinking about your own thinking: knowing what you know, and more importantly, knowing what you don't. It's the voice that says *wait, I'm not sure that's right.*

Here's the trap. To notice a grammar mistake, you need to know the grammar rule. To notice a weak argument, you need to understand what a strong one looks like. To know your joke didn't land, you need a decent sense of what's funny. The skill and the self-check are built from the same parts.

So beginners don't experience themselves as beginners. They experience a world that seems simpler than everyone keeps saying it is.

The most interesting part of the 1999 paper is what happened at the *other* end:

• **Top performers underestimated themselves.** They scored around the 87th percentile but guessed lower, because they assumed if the test felt easy to them, it must have felt easy to everyone. That's a version of the **false consensus effect**.
• **Top performers recalibrated fast.** When shown other students' answers, the high scorers raised their self-estimates. The low scorers mostly didn't, because they couldn't tell which answers were better.
• **Training fixed the blind spot.** In the fourth study, a short lesson in logic didn't just improve scores. It made low performers more accurate about how poorly they had done before. Learning the skill installed the mirror.

That last finding is the hopeful part. The Dunning-Kruger effect isn't a permanent personality type. It's a stage that shows up in whatever area you're newest in, and good feedback is the way out.

The Dunning-Kruger effect is real but smaller and more nuanced than internet memes suggest. Researchers including Edward Nuhfer and colleagues (2016) and Gilles Gignac and Marcin Zajenkowski (2020) showed that part of the classic chart comes from regression to the mean and the near-universal better-than-average effect, while the core finding, that low performers misjudge themselves most, still holds.

If you've seen the Dunning-Kruger effect online, you've probably seen a curve with a peak called "Mount Stupid," a dip called the "Valley of Despair," and a slow climb up the "Slope of Enlightenment." It's a great meme. It's also not in the original paper. Kruger and Dunning never plotted confidence over time or claimed beginners start overconfident and then crash.

Serious critiques of the effect fall into two buckets:

• **The better-than-average effect.** Most people rate themselves above average on almost everything, from driving to kindness. If *everyone* guesses roughly the 60th to 70th percentile, the people who actually scored at the bottom will automatically look the most overconfident.
• **Regression to the mean.** Test scores contain some noise. When you sort people by their score and compare that with a separate, also-noisy self-estimate, the two will drift toward each other at the extremes even with no psychological effect at all. Nuhfer's team demonstrated this by generating random data that produced a Dunning-Kruger-shaped chart.

Gignac and Zajenkowski used statistical methods designed to avoid those traps on a sample of nearly 1,000 people rating their own intelligence. They still found the effect, just a much weaker one than the original chart implies.

Dunning himself has acknowledged the debate while arguing the central idea survives: people are often poor judges of their own competence, and the gap tends to be widest where skill is lowest.

So the honest version sounds less like a punchline and more like a warning label: *everyone* is a little overconfident, and the less experience you have in an area, the less you should trust your gut about how you're doing.

The loudest friend seems smartest because groups read confidence as competence. In a 2009 study, Cameron Anderson and Gavin Kilduff found that people with dominant personalities gained influence in problem-solving groups by speaking first, speaking often, and sounding certain, even though they weren't more accurate. Their teammates sincerely perceived them as more capable.

Anderson and Kilduff put students into small groups to solve math problems together. Afterward, group members and outside observers rated who was most competent. The dominant participants were consistently rated as the most capable. Their actual math scores? Not meaningfully better.

What made them look smart were behaviors that *signal* competence:

• Answering first
• Offering more answers overall
• Speaking in a calm, sure, factual tone
• Not hedging with "I think" or "maybe"

Anderson later found, with Sebastien Brion, Don Moore, and Jessica Kennedy (2012), that overconfident people tend to gain status in groups, partly because their certainty reads as skill.

Now open your group chat. The person who replies in eight seconds with "no it's definitely Thursday" gets believed over the person who replies ten minutes later with "I could be wrong, but I think the email said Friday?" The second friend checked the email. The first friend sounded like they did.

This is where Dunning-Kruger and group dynamics team up. The less someone knows about a topic, the fewer reasons they see to hedge, so the more certain they sound, so the more influence they get. Meanwhile, the expert's honest uncertainty comes across as weakness.

None of this makes the confident friend a villain. Most of the time they genuinely believe what they're saying, and the group often *wants* someone to make a call. The problem only shows up when nobody can tell the difference between sure and right.

Being confident isn't a bad thing. Confidence gets groups moving, makes decisions possible, and helps ideas get heard, while doubt catches mistakes and keeps plans grounded. The Dunning-Kruger problem isn't confidence itself but confidence that's out of sync with skill, which can happen at either end: overconfident beginners and underconfident experts.

It would be easy to turn this guide into "loud people bad, quiet people wise." That's not what the research says, and it's not how friend groups actually work.

The friend who answers first is often the reason anything happens at all. They pick the restaurant, book the trip, and say what everyone was thinking. The friend who hesitates is often the reason the trip doesn't go off the rails. Groups need both.

And the mirror image of Dunning-Kruger is just as real. Plenty of highly skilled people sit at the back of the chat assuming everyone knows what they know, and some carry full-blown **impostor feelings**, a pattern psychologists Pauline Clance and Suzanne Imes described in 1978. Overconfidence and underconfidence are two flavors of the same miscalibration.

On the vibe check, this tension shows up across several sliders where both ends are compliments:

• **Smooth Operator ↔ Straight Shooter** (how you handle the truth): Straight Shooters cut through the noise. Smooth Operators make the truth land without bruising anyone.
• **Slow Burn ↔ Live Wire** (where your energy comes from): Live Wires fill the room. Slow Burns often do the quieter, deeper thinking.
• **The Anchor ↔ The Daredevil** (how you meet risk): Daredevils leap. Anchors make sure there's ground to land on.
• **The Director ↔ Main Character** (where you stand in the room): Main Characters are the show. Directors make the show actually work from backstage.

The Dunning-Kruger moment isn't living at one end of any slider. It's when the end you *think* you're on and the end your friends see you on are miles apart.

Trimmed-mean anonymous consensus calibrates overconfidence by replacing one person's self-image with a combined view from at least three friends. Because ratings are anonymous, nobody has to soften them. Because results use a median for small groups and a 10%-trimmed mean at five or more, one harsh or overly generous rating can't swing the outcome. The gap between self and group becomes the feedback.

Remember Kruger and Dunning's fourth study: the thing that fixed the blind spot was feedback. Not a lecture, not a roast. Information the person couldn't generate on their own.

The trouble is that real-life feedback about confidence is almost impossible to give. Telling a friend "you sound way more sure than you should be" is awkward, risks the friendship, and usually triggers the exact doubling-down you were hoping to avoid. So people stay quiet, and the gap grows.

Organizational psychologist Tasha Eurich's research (2017) found that most people believe they're self-aware, but only a small minority actually are. She separates **internal self-awareness**, how well you understand yourself, from **external self-awareness**, how well you understand how others see you. Dunning-Kruger is fundamentally an external self-awareness problem, which means it can't be solved from inside your own head.

Here's how to run the check on [Am I like this?](/new):

• **Start the vibe check.** Rate yourself on all 6 axes: Candor, Order, Voltage, Compass, Leap, and Orbit.
• **Drop the anonymous link in the group chat.** You need at least **three responses** to unlock results, so no single friend can be identified or blamed for a rating.
• **Let the math stay fair.** With three or four responses, each axis uses the **median**. At five or more, it switches to a **10%-trimmed mean**, cutting the most extreme ratings at both ends so one outlier, harsh or flattering, can't distort your shape.
• **Read the gaps, not the grades.** Your radar overlays your self-rating with the group's view. Big gaps are the interesting part.

What to look for:

• **You rated yourself a Straight Shooter, but friends see a Smooth Operator?** You might be less blunt than you think. Or more diplomatic than you give yourself credit for.
• **You placed yourself backstage as a Director, but friends say Main Character?** Your presence may carry more weight in the room than you realize, which is worth knowing before you share your next hot take.
• **Your self-image and the group's view basically match?** That's calibration. It's rarer than it sounds and genuinely worth celebrating.

The best part is that it works in both directions. The overconfident friend sees, gently and privately, where their certainty outruns the room's impression. The underconfident friend sees that everyone else noticed the competence they keep downplaying. Same check, opposite lesson, zero roasting required.