Confidence in one's knowledge can be assessed with questions of the following kind:
Which city has more inhabitants?
How confident are you that your answer is correct?
50%, 60%, 70%, 80%, 90%, 100%
If you answer 50%, then you are guessing. If you answer 100%, then you are absolutely sure of your answer.
Two decades of research into this topic has demonstrated that in all cases where subjects said they were 100% certain of an answer, the relative frequency of correct answers was 80%. Where subjects said they were 90% certain of an answer, the relative frequency of correct answers was about 75%. If subjects said they were 80% confident of an answer, the relative frequency of correct answers was in fact 65%, and so on.
In other words, values for confidence were systematically higher than relative frequencies. This is the overconfidence bias which has been demonstrated using a variety of tasks, including those that are impossible or nearly impossible. Examples include predicting the winners in 6-furlong horse races or diagnosing the malignancy of ulcers.
Overconfidence also occurs when accumulating evidence, for example case-study material about some demonstrable effect or other, from which certain predictions are then made. There is a point in the information-gathering process when predictive accuracy reaches a ceiling. Nevertheless, confidence in one's conclusions continues to rise as more information is received. Towards the end of the information-gathering process, most judges are overconfident about their judgements.