grading-level-division-reference-calibrated

To calibrate grade boundaries, align grading boundaries with reference score distributions and rubric descriptors to mitigate year-to-year bias. The process applies anti-bias adjustments to match current cohorts with reference-year expectations, generating a deterministic level division file.

Reference-calibrated level division computes score-to-level mapping using empirical data from reference students combined with rubric means. It applies a bias prevention protocol before finalizing boundaries, ensuring grade levels reflect actual rubric performance rather than arbitrary cutoffs.

Prevent year-to-year grading bias by applying anti-bias adjustments that align current cohort scores with reference-year expectations. This protocol compares current distributions against reference scores and rubric descriptors before finalizing grade boundaries.

Generating a deterministic level division requires final_scores.json, rubric reference_scores.json, and reference_calibration.md files. These inputs provide the score distributions and rubric descriptors needed to compute boundaries and emit per-student final levels.

Yes, reference score distributions determine initial level boundaries from empirical mapping and rubric means. The system applies anti-bias adjustments to align different cohorts with reference-year expectations, ensuring consistent grading standards across varying student populations.

Adjustment notes document the anti-bias adjustments applied to align current scores with reference-year expectations. These notes accompany the per-student final levels in the output, providing transparency about boundary modifications made during the calibration process.