What Is the Hard-Easy Effect?

📚
Easy Tasks
You're 90% accurate
but only 70% confident
→ Underconfident
⟷
🧩
Hard Tasks
You're 30% accurate
but 60% confident
→ Overconfident

Why Does This Happen?

  • Regression to the mean: Confidence regresses toward 50%, while accuracy varies more widely
  • Anchoring: We anchor on a default confidence level and adjust insufficiently
  • Scale compression: We use the confidence scale suboptimally at the extremes
  • Confirmation bias: We notice evidence for our answer more than against it

Not the same as Dunning-Kruger! D-K is about skill level (unskilled people don't know they're unskilled). The Hard-Easy Effect is about task difficulty—even experts show this pattern on hard questions in their own field.

Key Research

Lichtenstein & Fischhoff (1977)
Foundational study: People were 65-70% confident when only 50% correct on hard questions. On easy questions, confidence lagged behind accuracy.
Gigerenzer et al. (1991)
The "overconfidence bias" largely disappears when questions are randomly sampled from natural environments rather than selected to be difficult.
Moore & Healy (2008)
Distinguished three types of overconfidence: overestimation (of your performance), overplacement (vs. others), and overprecision (in your beliefs).
Brenner et al. (1996)
Weather forecasters and bridge players who receive regular feedback show excellent calibration—the Hard-Easy Effect can be reduced with practice and feedback.

Real-World Implications

Financial Decisions

Investors are overconfident about complex, hard-to-predict markets and underconfident about simpler decisions. This leads to excessive trading and poor portfolio allocation.

Medical Diagnosis

Doctors show overconfidence on rare, difficult diagnoses and appropriate confidence on common conditions. Patients need to understand diagnostic uncertainty.

Legal Judgments

Eyewitnesses are often overconfident about difficult identifications. Confidence doesn't reliably predict accuracy for hard cases.

Project Planning

Teams are overconfident about complex, uncertain projects and underconfident about routine tasks. This distorts resource allocation.

"The confidence people express reflects not only what they know but also what they don't know about what they don't know."
— Lichtenstein, Fischhoff & Phillips (1982)

How to Improve Your Calibration

1. Seek Feedback

Professionals who receive regular, timely feedback (weather forecasters, bridge players) show near-perfect calibration. Track your predictions and outcomes.

2. Consider Alternatives

Before committing to an answer, genuinely consider why you might be wrong. Generate reasons the alternative could be correct.

3. Use Base Rates

On hard questions, remember that most people (including you) will be wrong. Anchor on typical accuracy rates for similar questions.

4. Calibration Training

Practice estimating probabilities and getting feedback. The skill transfers across domains once learned.

The Goal: When you say you're 80% confident, you should be right about 80% of the time. Not more, not less. This is perfect calibration.