Overconfidence and Prediction

The Confidence Calibration Problem

Ask people trivia questions and have them rate their confidence. When they say "90% sure," they're right about 70-75% of the time. When they say "99% sure," they're right about 85% of the time.

Our confidence consistently exceeds our accuracy.

This isn't occasional error. It's systematic overconfidence.

Three Types of Overconfidence

Overestimation

We overestimate our own abilities, knowledge, and performance.

Examples:

  • How well did you perform? (Better than you actually did.)
  • How much do you know about this topic? (More than you actually do.)
  • How likely are you to succeed? (More likely than you actually are.)

Overplacement

We overestimate our relative standing compared to others.

Examples:

  • 90% of drivers consider themselves above-average drivers.
  • Most students think they're smarter than average.
  • Most entrepreneurs believe their business will beat the odds.

This is mathematically impossible. Not everyone can be above average.

Overprecision

We're too certain in our beliefs. Our confidence intervals are too narrow.

Example: "I'm 90% confident the answer is between X and Y." But the true answer falls outside that range far more than 10% of the time.

Overprecision is perhaps the most consequential form — it leads us to believe we know things we don't.

Where Overconfidence Hurts

Planning Fallacy

We underestimate how long tasks will take, how much they'll cost, and what can go wrong.

Classic findings:

  • Construction projects average 50% over budget
  • Software projects take 2-3x longer than estimated
  • Personal tasks take 40% longer than expected

We focus on the ideal scenario and neglect the realistic one.

Prediction

We make confident predictions that turn out wrong.

Expert predictions are often no better than random, yet experts remain confident.

Political forecasts, economic forecasts, technology predictions — the track record is poor, yet confidence stays high.

Entry Decisions

Too many businesses enter markets expecting to succeed where most fail.

If the average new restaurant fails within three years, but every new restaurateur believes they'll beat the odds, the problem is clear.

Investment

Overconfident investors:

  • Trade too much (hurting returns through costs and mistiming)
  • Under-diversify (too confident in their picks)
  • Stick with losers (confident it'll turn around)

Why Overconfidence Exists

Evolutionary Advantage

Confidence may provide social and competitive advantages — even unjustified confidence. Leaders seem more capable when confident. Overconfidence might have been adaptive.

Limited Feedback

We rarely get clear, immediate feedback on our predictions. Failed forecasts are forgotten; hits are remembered.

Self-Serving Attribution

We attribute success to our skill and failure to bad luck. This preserves confidence despite evidence.

Hindsight Bias

After events, we misremember our predictions as better than they were. "I knew that would happen" — even when we didn't.

The Dunning-Kruger Effect

What It Is

The least competent people are often most overconfident about their abilities. They lack the knowledge to recognize their own lack of knowledge.

Meanwhile, highly competent people often underestimate their relative abilities — they know how much they don't know.

The Pattern

  • Beginners: High confidence, low competence
  • Learners: Declining confidence as they realize complexity
  • Competent: Moderate confidence, high competence
  • Experts: Appropriately calibrated (though still prone to overconfidence)

Improving Calibration

Track Your Predictions

Keep records of predictions with confidence levels. Review regularly. You'll discover your actual accuracy rates.

Widen Your Intervals

If you think the range is X to Y, deliberately expand it. Your initial range is almost certainly too narrow.

Consider Alternatives

Before committing to a belief, actively consider how it could be wrong. What would have to be true for the opposite to be correct?

Use Base Rates

Before estimating your chances of success, ask: "What's the success rate for people in similar situations?"

Don't assume you're special until you have evidence.

Seek Disconfirming Evidence

Actively look for evidence against your beliefs. If your belief survives genuine challenge, it's more trustworthy.

Pre-Mortem

Before starting a project, imagine it failed. What went wrong? This surfaces risks optimism obscures.

Check with Outside View

Your inside view focuses on your specific situation. The outside view asks: "How do similar cases usually turn out?"

Useful Overconfidence?

When It Might Help

Motivation: Believing you'll succeed motivates effort.

Persistence: Overconfidence keeps you going through early failures.

Signaling: Confidence impresses others, opens opportunities.

The Balance

Some optimism may be functional. But dramatic overconfidence leads to poor decisions, overextension, and preventable failures.

The goal: Realistic confidence that acknowledges uncertainty while still enabling action.

AI Prompt: Confidence Check

Help me calibrate my confidence in a prediction or belief.

My prediction/belief: [What you believe]
My confidence level: [How sure you are, as a percentage]

Help me:
1. Identify reasons my confidence might be too high
2. Consider alternative possibilities I might be ignoring
3. Find the base rate for similar predictions
4. Suggest what a well-calibrated confidence level might be
5. Identify what would prove me wrong

What's Next

Understanding biases is useful. But how do we design environments that make good decisions easier?

Next chapter: Choice architecture — designing environments for better decisions.