Self-grading usually means reading your answer and deciding it feels fine — a vibe check that passes almost everything. The answer looks complete because you just wrote it; the gaps are invisible because they're yours. A TA grading the same answer would mark what the vibe misses: the verb half-answered, the mechanism asserted instead of shown, the trap walked into.

The fix is a rubric with three boxes, applied a minute after the attempt. Box one: did the answer do the verb — if the prompt said "explain," did it explain or just name? Box two: did it give the mechanism — the causal step, the reason, not just the claim? Box three: did it name the trap — the wrong answer the question was built to catch? Anything under 3/3 isn't done; it becomes a new stem to rewrite.

Why vibe-grading fails

Calling an answer fine because it feels complete is recognition grading — the answer matches the shape of an answer, so it passes. But the exam grades content, not shape: did you explain the mechanism, or assert the conclusion? The vibe can't tell the difference because the vibe is measuring fluency — how smoothly the answer reads — and fluency is exactly what a plausible wrong answer has.

Scoring is the metacognition half of practice — the attempt is retrieval, the grade is feedback, and feedback is what keeps errors from fossilizing. Retrieval without correction just practices whatever you produced, including the mistakes; the three-box grade catches the specific kind of wrong — missing mechanism, unaddressed trap — and converts it into the next stem. The rubric is what turns a practice answer into a repaired one.

The three-box rubric

Box one: the verb. Every prompt has an operation — explain, define, compare, derive — and answers lose points by doing a different one. "Explain why" answered with a definition is a miss even if the definition is right: the verb wasn't done. Grade the operation first because it's the most common silent failure — right content, wrong verb.

Box two: the mechanism. The difference between a claim and an answer is the causal step — the "because" that shows the working, not just the result. "There is a shortage" is a claim; "the ceiling holds P below equilibrium, so Qd exceeds Qs and time-in-line does the rationing" is a mechanism. The mechanism is what partial credit pays for and what the exam's harder questions require.

Box three: the trap. Most exam questions are built around a wrong answer that looks right — the intuitive-but-false path. A complete answer names the trap: not just the right mechanism but the wrong one it replaces. "The trap is blaming greed instead of the price wedge" shows you know what the question was testing, which is worth points and proves understanding.

Anything below 3/3 gets rewritten, not reread. The fix for a missing mechanism is a new attempt with the mechanism in it — a rewrite, scored again. Rereading the model answer feels like repair but produces the same answer next time; the rewrite is the rep that changes what you produce.

A 40-minute session on the clock

The weak grade passes a slogan — "shortage is the reason" restates the conclusion without the causal chain, which is the mechanism the verb "explain" demanded. The strong grade catches all three boxes: the verb half-done, the mechanism absent, the trap unnamed. The rewrite then shows what full marks look like — the wedge, the quantity gap, the rationing — and names the wrong path the question was built to punish.

Why the one-minute wait matters

Grade a minute after the attempt, not during — the answer is still warm, and warm answers grade themselves leniently because you remember what you meant. The minute of distance is what lets you read what's actually on the page instead of what you intended to write. A TA doesn't know what you meant; the rubric applied cold is the closest you get to reading like one.

The wait also separates the two skills. Attempting is retrieval; grading is evaluation, and doing them simultaneously means the grader is biased by the retriever's memory of trying. The pause makes the grade about the answer, not the effort — which is the only grade that predicts the exam's.