Analytic Rubric Builder — Observable, Defensible Descriptors
Build an analytic rubric with observable, calibration-tested descriptors for any assignment, plus a student-friendly self-check version. Designed so two teachers grade the same work consistently.
Variables
You're an assessment specialist. Build an analytic rubric where two teachers grading the same work land within one level of each other. Every descriptor is observable — what you'd actually SEE in the work — not "good" vs. "excellent." Grade: {{grade_level}} | Subject: {{subject}} Assignment type: {{assignment_type}} What it's supposed to demonstrate (objective/standard): {{objective}} Criteria to assess (or ask me to propose them): {{criteria}} Levels: {{{levels: 4 (Exceeding/Meeting/Approaching/Beginning), 3 (Advanced/Proficient/Developing), 4-3-2-1 points}}} Grading use: {{{use: formative feedback, summative grade, student self-assessment, peer review}}} **Deliver:** **RUBRIC TABLE** Rows = criteria (3-5, each assessing one distinct thing). Columns = levels. Each cell is a specific, observable descriptor. The jump between levels is a concrete, nameable difference (frequency, accuracy, depth, independence) — not a vaguer or stronger adjective. Weak: "Good use of evidence" then "Excellent use of evidence" Strong: "Uses 1-2 pieces of evidence; some support the claim" then "Uses 3+ pieces of relevant evidence; each clearly supports the claim" **POINT VALUES (if grading)** How levels map to points and how criteria weight against each other. Flag the criterion that should carry the most weight and why. **STUDENT-FRIENDLY VERSION** The same rubric in plain language students can self-check against before submitting. First person ("I included..."). **EXEMPLAR NOTES** 2-3 sentences describing what top-level work looks like for this specific assignment — the picture students should aim for. **CALIBRATION CHECK** 2-3 borderline cases (e.g., "strong evidence but disorganized") and how to score them, so grading stays consistent across a stack of papers. Rules: Every descriptor is observable in the work itself — no mind-reading, no comparing to other students. Level differences are concrete (count, accuracy, depth), not adjective inflation. Criteria don't overlap. Share the rubric with students before they start. The plain-language version uses words the grade level actually understands.
