Observation-journal defense
This is a live exam over a record the student built themselves. Over weeks, they keep a dated sky-observation journal at home — moon phases night after night, a planet creeping against the stars, sunspots, the turning constellations. Then they bring it in and defend it. The guide starts asking: which nights are these, what pattern runs through them, what geometry makes that pattern, and where the honest line falls between what was seen and what was inferred. There is no figure to look up and nothing to copy: a personal, time-stamped record can’t be generated the night before, and the student stands over their own pages and defends the run out loud.
| Criterion | Not yet | Approaching | Mastered |
|---|---|---|---|
| Consistency & dating of entries | Entries are sparse or undated, or the whole journal was clearly backfilled in one sitting. | Most entries are dated, but gaps break the run or the dates don't line up with the sky. | Keeps dated, regular entries across weeks with no suspicious gaps — a genuine over-time record, not a cram. |
| Quality of sketches & records | Vague scribbles that don't capture what was actually seen. | Recognizable sketches, but missing the time, the direction, or the detail that makes them useful. | Clear, labeled sketches carrying date, time, and sky position — the Moon's terminator, a planet's place among the stars, a sunspot group. |
| Correctness of identified patterns | Finds no pattern, or names one the entries do not actually show. | Spots a pattern — the lunar cycle — but misreads its shape or its timing. | Reads the pattern correctly from their own entries: the waxing–waning cycle, a planet's slow drift (and its retrograde-ish loop), or the seasonal turn of the constellations. |
| Honest inference vs. overreach | Claims more than the record supports, or invents detail to fill a gap. | Makes reasonable inferences but blurs what was seen with what was assumed. | Separates observation from inference cleanly, and won't claim a conclusion the journal can't carry. |
| Oral defense under questioning | Folds at the first follow-up or recites a memorized line that doesn't fit the journal. | Answers some follow-ups, falters when asked to justify a pattern or a date. | Handles unrehearsed follow-ups about their own record with sound, on-the-spot reasoning. |
“These dates run three weeks straight, so you can watch the terminator march across the Moon and the lit fraction grow — that’s the waxing half of the cycle. And Mars slid a little west against the same background stars over these nights before it turned back east; that’s the retrograde loop, not the Moon moving. I only sketched what I actually saw — anything I’m guessing at, I marked as a guess.”
“It got bigger, I think? I did most of these the night before it was due. Something moved across the sky but I’m not sure which one.”
This assessment is AI-proof by design: it is a personal, dated record built over weeks and defended in person. No chatbot can keep a month of your own sky nights, narrate the decisions behind sketches it never made, or hold up under a follow-up about a pattern only your entries show. Mastery is shown by keeping the record and defending it — not by submitting.