Dunning–Kruger Effect, explained.
The Dunning–Kruger effect concerns patterns of self-assessment and performance: recognizing mistakes can itself require knowledge of the task being judged.
Why it happens
Being confident and being correct are distinct. Someone may lack the information needed both to answer well and to recognize an error. But statistical patterns also depend on measurement, noise and task design, so a popular cartoon is not a universal learning trajectory.
Dunning and Kruger's research examined how task performance and awareness of mistakes can relate to self-assessment. Confidence and competence are different measurements.
Read the result
Record confidence before each answer is revealed. Compare average confidence with accuracy after five questions, then treat the difference as a small-sample observation. The exercise does not establish a Dunning–Kruger effect in you.
A worked example
Checking forecast calibration
A team records confidence for each forecast before seeing what happened.
Across many comparable forecasts made with 80% confidence, it checks whether roughly 80% were correct.
Feedback can reveal overconfidence or underconfidence in that task. A few misses cannot establish a person's general competence.
OPTIONAL DEEPER DETAILGo deeper: inside the model
Inside this model
Five fixed, two-option questions ask for confidence from 50% to 100% before feedback. The summary shows your accuracy, average stated confidence and their difference. It does not estimate your percentile or produce the popular confidence-versus-expertise cartoon.
Where this idea is useful
A practical use
After a prediction, record how sure you were before learning the outcome. Over many comparable decisions, check whether confident judgments are correct as often as expected.
A common misconception
“High confidence proves someone has little knowledge.”
Experts can be appropriately confident, and novices can be uncertain. Confidence alone does not identify performance or the research effect.
What this explanation leaves out
- Five questions cannot establish a population effect or diagnose a person's competence. Item difficulty, prior knowledge and chance affect this tiny sample, and explanations of the original statistical pattern remain debated.
Is this the familiar mountain-shaped confidence curve?
No. The exercise records actual answers and stated confidence. The well-known cartoon should not be treated as the original study's measured universal curve.
What feedback would help you learn whether your confidence matches your accuracy?
Associated thinkers
Further reading
Explore the original research or the teaching reference behind this experiment.