Why certainty feels so convincing

We tend to assume that confidence is produced by knowledge. Sometimes it is. Sometimes it is produced by personality, incentives, incomplete information or the pleasant absence of anyone checking.

That distinction matters because confident people shape decisions. They win attention, reassure colleagues and make uncertain plans sound manageable. Accuracy arrives later, by which point the budget has been approved and everyone has a branded notebook.

The central problem is not confidence itself. It is poor calibration: a mismatch between how certain we feel and how often we are actually right.

Confidence is a forecast

If someone is well calibrated, statements made with 80% confidence should prove correct approximately 80% of the time. They will still be wrong occasionally. That is what the other 20% is doing there.

Most people do not naturally keep this kind of score. We remember being right, explain away being wrong and quietly edit the confidence we originally held. A prediction that felt certain can later become something we were merely ‘considering.’ Memory offers a surprisingly flexible audit function.

This is why confidence needs to be treated as a forecast rather than a performance. It should be recorded, tested and improved.

Why intelligent people still get it wrong

Overconfidence is often presented as an arrogance problem. That is too convenient.

Accurate self-assessment requires mental work. You must consider the evidence supporting your view, the evidence against it, the reliability of your sources and the possibilities you have missed. Under pressure, the brain replaces difficult questions with easier ones.

Instead of asking, ‘How strong is the evidence?’ we ask, ‘Can I explain this clearly?’

Instead of asking, ‘Is this candidate capable?’ we ask, ‘Did this candidate seem confident?’

Instead of asking, ‘Is the strategy working?’ we ask, ‘Are the headline numbers going up?’

These substitutions are efficient, plausible and occasionally ruinous.

The popular version of the Dunning–Kruger effect adds another oversimplification. The familiar Mount Stupid graph was not produced by the original research. The underlying finding was that poor performers showed a larger gap between estimated and actual performance, while stronger performers were generally closer to reality.

That does not mean beginners are uniquely foolish. Self-evaluation is simply another skill, and people can lack both the ability to perform a task and the ability to recognize their mistakes.

Business provides noisy feedback

In a controlled environment, people can learn from clear outcomes. Chess players receive prompt feedback: the move works or it does not.

Organizations operate in messier conditions. Markets move, competitors react and good decisions can produce bad outcomes. Reckless decisions can also succeed, which is more dangerous because success tends to end the investigation.

Nick Leeson’s unauthorized trading at Barings Bank illustrates the problem. An early hidden loss was eventually recovered after he took increasingly risky positions. The recovery did not prove that his method was sound. It rewarded him for surviving it.

When another loss appeared, he repeated the behavior. His positions grew while management continued sending money, reassured by apparent profits and Leeson’s reputation. The bank’s controls failed to challenge the story until the losses became fatal.

This was not confidence operating alone. It was confidence combined with weak oversight, complex information and unreliable feedback.

Confidence is socially useful

Overconfidence persists partly because it can improve a person’s status.

Confident people speak sooner and make cleaner recommendations. They appear decisive. In meetings, recruitment and leadership contests, this creates an advantage before accuracy can be tested.

Organizations therefore face a design problem. If certainty is rewarded more quickly than judgment, people learn to display certainty. The culture may claim to value truth while promoting the person least inconvenienced by doubt.

The answer is not to fill every meeting with nervous disclaimers. Businesses need people who can act under uncertainty. But action and certainty are not the same thing.

A capable decision-maker can say, ‘I believe this is the best option, with roughly 70% confidence. These are the assumptions, and this is what would change my view.’

That is not indecision. It is risk management expressed in complete sentences.

How to improve calibration

Start by recording important predictions. Give them probabilities and review the outcomes. Over time, you will discover whether your 80% means 80%, 55% or ‘I would strongly prefer this to happen.’

Write down the assumptions behind major decisions. Invite someone to identify missing information rather than merely argue the opposite case. Reward people who update their views when evidence changes.

Finally, review near misses and lucky successes. A favorable outcome does not automatically validate the process that produced it.

Confidence is valuable when it reflects evidence. When it does not, it is merely a persuasive feeling with access to the budget.

Further reading

• Unskilled and Unaware of It (Journal of Personality and Social Psychology via PubMed) - The original Kruger and Dunning paper on performance and inaccurate self-assessment.
• The Dunning–Kruger Effect Revisited (Nature Human Behaviour) - Examines the mechanisms and statistical questions behind the well-known effect.
• Known Unknowns: A Critical Determinant of Confidence and Calibration (Management Science) - Shows why considering missing information can be more useful than merely arguing the opposite case.
• Report into the Collapse of Barings Bank (UK Board of Banking Supervision via GOV.UK) - The official inquiry into the controls, supervision and management failures behind Barings’ collapse.
• The Challenger Accident: Flaws in the Decision-Making Process (NASA) - The Rogers Commission’s account of incomplete information, communication failures and managerial judgment.