Ask how sure they are, not just what they answered
A right answer at 50% confidence and a wrong answer at 95% confidence are both news. The answer alone hides both.
Aug 18, 2026 · 1 min read
Students are reliably poor judges of what they know: rereading and familiarity create an illusion of competence that a confidence rating exposes (Koriat & Bjork, 2005). Pair any answer with a quick "how sure are you?" and you get a second layer of data: the confidently wrong students need a misconception repaired, while the hesitantly right ones need consolidation, and those are different lessons. (The confidence-calibration-check skill in Manning's education-agent-skills library builds a whole routine around this.)
In practice
- Keep the scale tiny: sure / think so / guessing. Held-up fingers work, or a letter written next to the answer.
- Hunt the two interesting corners: confident and wrong (address it publicly, it's rarely one student), hesitant and right (more practice, not more explanation).
- Before a test, have students predict their score per topic, then compare after. The gap is the study plan.
Quick takeaway: correctness tells you about the answer. Confidence tells you about the student.