People with low competence in a domain significantly overestimating their own ability, because they lack the skill to see the gap.
Imagine you just learned three chords on guitar and you think you sound amazing because you don't know enough about music to hear what a real guitarist sounds like. Meanwhile, the actual expert guitarist thinks they're only okay because they compare themselves to even better players. You need skill to recognize skill — so when you don't have it, you can't even tell what you're missing.
The Dunning-Kruger effect describes a metacognitive failure in which individuals who perform poorly on a task lack the very skills needed to recognize how poorly they have performed, creating a 'dual burden' of incompetence and unawareness. This leads bottom-quartile performers to place themselves well above the median in self-assessments, while top-quartile performers slightly underestimate their relative standing because they assume the task was equally easy for everyone. The effect is domain-specific rather than a statement about general intelligence: a person can be well-calibrated in one area yet dramatically overconfident in another where they lack expertise. Notably, when low performers are trained and become more competent, their self-assessments become more accurate, as they finally acquire the metacognitive tools to recognize quality work.
The same glitch looks different depending on the terrain. Finance, medicine, a relationship, a team — same mechanism, different costume.
Novice investors who experienced a bull market early in their investing careers often dramatically overestimate their stock-picking ability, attributing market-wide gains to personal skill and taking on excessive risk in subsequent trades.
Patients who research symptoms online may become convinced they understand their condition better than their physician, rejecting diagnostic recommendations and pursuing alternative treatments based on superficial understanding.
Justin Kruger and David Dunning, Cornell University, 1999. Published as 'Unskilled and Unaware of It: How Difficulties in Recognizing One's Own Incompetence Lead to Inflated Self-Assessments' in the Journal of Personality and Social Psychology, 77(6), 1121-1134.
In ancestral environments, overconfidence served as a social signaling mechanism that could confer status, attract mates, and deter competitors without the cost of actual physical confrontation. Individuals who projected confidence — even unwarranted confidence — were more likely to gain leadership roles, resources, and reproductive opportunities in small tribal groups where objective performance metrics were rarely available.
LLMs exhibit a Dunning-Kruger-like pattern: they generate confident, fluent text even on topics where their training data is sparse or contradictory, producing plausible-sounding but incorrect information (hallucinations) without any self-awareness of their limitations. Research has also shown that AI tools amplify the effect in users — people using ChatGPT to complete tasks consistently overestimate their own performance, with AI-literate users paradoxically showing the greatest overconfidence. Additionally, LLMs tested on coding tasks in rare programming languages show stronger miscalibration between reported confidence and actual accuracy, mirroring the human pattern.
This page taught you the name. The deck turns the name into reflex. 1,100+ swipeable scenarios, 1,100+ defenses, 650+ detection prompts — spaced-repetition Swipe Deck, unlimited Spot-the-Bias Quiz, Defense Playbook, Pre-Flight, My Blindspots, Cheat Sheets, Field Guide e-book. $39.53$59.
