Explicit capstone instruction
A useful experiment or simulation changes one relevant feature at a time when possible and records raw evidence.
Define independent, dependent, and controlled variables before testing.
Plan repeated trials, calibration, and a revision criterion so the capstone does not become a one-shot demonstration.
Vocabulary: test plan; independent variable; control; trial; calibration.
Worked example
If model A predicts 10 units and model B predicts 16 units, a test with ±5-unit uncertainty cannot discriminate well; improving measurement precision may be more valuable than adding complexity.
Guided question
How will your test decide whether the model is good enough for the stated criterion?
Independent work
- Define variables and controls.
- Choose raw measurements.
- Specify trial count.
- Write a revision trigger.
Misconception check
A test that cannot distinguish competing outcomes is weak even if it produces data.
Formative check
What should a revision trigger be tied to?
Reveal response
A measurable criterion, model residual, or repeatable test failure.
Mentor evidence conference
Ask Archaemenes to challenge system boundaries, predictions, uncertainty, and whether revisions follow the evidence.
Teacher guidance
Foundation: Use a test-plan checklist.
Core: Design a reproducible comparison.
Extended: Optimize information gained per trial.