Grade the first three papers blind
Cover the names. Your first three set the scale for the whole stack, and that is exactly where bias does the most damage.
Aug 2, 2026 · 1 min read
The first few pieces you mark calibrate your internal standard for everything after them, and if you know whose work you're reading, expectation leaks into the calibration. This is measured, not hypothetical: a meta-analysis of 20 grading experiments found a significant bias effect when markers knew irrelevant things about the student, from prior performance to ethnicity to attractiveness (Malouff & Thorsteinsson, 2016). Anchor the scale on the work, not the reputation.
In practice
- Fold the name over, or open submissions in a random order with names hidden, for at least the first three.
- Mark one strong, one middling, one weak early if you can find them. A spread calibrates better than three of a kind.
- Re-read your first three after finishing the stack. If your later marking drifted, they're the ones to adjust.
- Digital submissions make this nearly free: sort by anything other than name.
Quick takeaway: calibrate on work, not on reputations.