The rock & mineral ID defense
This is a live exam at the bench. The student is handed unknown rock and mineral specimens and must identify each one — running a streak, testing hardness against a Mohs kit, checking cleavage against fracture under the hand lens, reading luster and crystal habit, and dropping dilute acid to see whether a carbonate fizzes. Working a dichotomous key, the student lets each result decide the next test. Then the guide starts asking: why call that cleavage and not fracture, what the streak rules out, why the key branches where it does. There is no worksheet to copy and no key to look up the answer in: the student stands over the tray and defends each identification out loud.
| Criterion | Not yet | Approaching | Mastered |
|---|---|---|---|
| Property-test technique | Performs tests carelessly — drags the wrong way on the streak plate, misjudges hardness, or floods a specimen with acid. | Runs the common tests (streak, hardness) but is inconsistent on cleavage vs. fracture, or applies the acid drop without control. | Performs each test with correct technique — streak on unglazed porcelain, Mohs hardness against the kit, cleavage vs. fracture under the lens, luster and habit, and a single controlled dilute-acid drop to check for fizz. |
| Working the dichotomous key | Ignores the key or jumps to a name without following its branches. | Follows the key but takes a branch the evidence does not support, or backtracks aimlessly. | Works the key couplet by couplet, letting each property result decide the next branch until one identity remains. |
| Evidence-based justification | Names a specimen with no property to back it up. | Cites one property but cannot chain several into a single call. | Justifies each identification by chaining streak, hardness, cleavage, luster, and the carbonate test into one conclusion. |
| Handling a wrong turn | Freezes or guesses when a test contradicts an earlier branch. | Notices the contradiction but cannot say which test to trust. | Recognizes a wrong turn, re-tests the doubtful property, and revises the identification with a clear reason. |
| Recording & oral defense | Records nothing usable and folds at the first follow-up. | Records most results but falters when asked to justify a call under pressure. | Records each property in the lab notebook and defends every identification aloud under unrehearsed questioning. |
“It left a pale streak, scratched glass but not quartz, so its hardness is about six, and it broke along two flat cleavage planes meeting near ninety degrees — that’s a feldspar, not quartz. No fizz with dilute acid, so it isn’t a carbonate either.”
“It’s shiny and kind of hard, so maybe quartz? I didn’t really run the streak. I’d have to guess which one without the answer key.”
This assessment is AI-proof by design: it happens at the bench, with real specimens and a real key, in real time. No chatbot can drag a mineral across a streak plate, feel a cleavage break, or watch for the fizz of dilute acid on a carbonate. The specimens differ from student to student — mastery is shown by identifying and defending in person, not by submitting.