Assessment Criteria
The assessment measures how the student thought, not how polished the text is. Five criteria, each assessed at four assessment levels. The methodological locks (The human anchor first, Honest Oracle, convergence ≠ corroboration) carry the most weight.
| Criterion | Excellent | Good | In Progress | Unsatisfactory | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1. The human anchor (written without a machine, honest baseline) | Clear, personal, and thoughtful; actively used for comparison in Step 3 | Present and utilized | Superficial or minimally used | Missing or obviously written after the fact | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 2. Three-level classification (strength of evidence) | Each claim correctly classified, Level 1 with a named source | Most correctly classified | Classification inconsistent | No classification — answers accepted as truth | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 3. Convergence ≠ corroboration | Clearly points out claims that appeared multi-verified but originate from a single source | Identifies at least one such instance | Mentions the concept without applying it | Considers consensus to be corroboration | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 4. Honest Oracle (resistance) | Detects hallucinations/name confusion, rejects ≥2 points with reasoning | Rejects with reasoning | Rejection is superficial | No criticism of own output | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 5. Control and accountability | Analyzes who controls the narrative + realistic data sovereignty/freedom actions + clear accountability statement | Actions and accountability present | Generally phrased | Missing action or accountability | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| (Fill-in template table) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Overall Grade (Guideline)
- Excellent: 4–5 criteria excellent, nothing below "Good".
- Good: most "Good" or better, methodological locks present.
- Pass: methodological locks present but shallow; at least "In Progress" everywhere.
- Fail: algorithmic output accepted as truth, no classification or no rejection.
Key rule: an empty "what is the counter-argument / what do I reject" section cannot receive higher than "In Progress". Resistance is not an afterthought — it is the assignment.