Overconfidence Effect
Type: Self-Assessment — Calibration Also Known As: Miscalibration, excessive certainty
Definition
Overestimating the accuracy of one’s beliefs, knowledge, or predictions. We think we know more than we do, are more skilled than we are, and can predict better than we can.
“I’m 95% confident the answer is between X and Y.” (Actual accuracy: 50%)
Three Types
- Overestimation: Believing we’re better than we are (90% of drivers say they’re above average)
- Overplacement: Believing we’re better than others (better than average effect)
- Overprecision: Being too certain about our knowledge (confidence intervals too narrow)
Examples
Example 1: Expert Predictions
Experts in politics, economics, and medicine routinely make predictions with high confidence that prove wrong. Expertise doesn’t eliminate overconfidence.
Problem: Expert status creates confidence beyond justified levels.
Example 2: Student Exam Predictions
Students predict they’ll score in the 70th percentile, then score in the 50th. The bottom quartile is most overconfident.
Problem: Dunning-Kruger effect — incompetence prevents recognizing incompetence.
Example 3: Entrepreneurial Ventures
Founders are wildly optimistic about success rates despite knowing 90% of startups fail. Each believes they’re the exception.
Problem: Overconfidence drives entry into unwinnable competitions.
Example 4: Confidence Intervals
When asked to give 90% confidence intervals, people’s ranges capture the true answer only 30-50% of the time. We’re systematically too certain.
Problem: We anchor on our best guess and adjust insufficiently.
Why It Happens
- Motivational benefit of self-confidence
- Confirmation bias — successes confirm ability, failures are explained away
- Illusion of control over random events
- Information gaps — we don’t know what we don’t know
- Social rewards for appearing confident
- Dunning-Kruger effect — skill is required to recognize lack of skill
How to Counter
- Calibration training: Practice estimating with feedback
- Consider the opposite: What would prove you wrong?
- Reference class forecasting: How did similar situations turn out?
- Pre-mortem: Assume you failed — what went wrong?
- Wider intervals: Deliberately expand confidence ranges
- Track predictions: Keep a record and score yourself
Dunning-Kruger Effect
The most unskilled people are most overconfident because:
- The skills needed to be competent are the same skills needed to recognize incompetence
- They don’t know what they don’t know
- A little knowledge is dangerous (creates false confidence)
Paradoxically, experts often show underconfidence in their domain because they understand the complexity.
Related Concepts
- Optimism Bias — Positive expectations about future outcomes
- Illusion of Control — Believing we influence uncontrollable events
- Hindsight Bias — Past overconfidence is forgotten
- Confirmation Bias — Confirms our self-assessment
References
- Moore, D.A. & Healy, P.J. (2008). The trouble with overconfidence
- Lichtenstein, S. et al. (1982). Calibration of probabilities: The state of the art
- Kruger, J. & Dunning, D. (1999). Unskilled and unaware of it
Part of the Convergence Protocol — Clear thinking for complex times.