Prepare a blank scoring rubric from an assignment prompt

Proposed inputs

1. Name

   Approved assignment

   Owner

   Educator

   Minimum scope

   One anonymized prompt and any required deliverable instructions.

   Quality check

   Include the version and distinguish required work from optional enrichment.

2. Name

   Scoring policy

   Owner

   Course lead

   Minimum scope

   Permitted criteria, desired total, weighting constraints and whether the rubric is formative.

   Quality check

   Resolve contradictions between the assignment and policy before drafting.

3. Name

   Curriculum evidence

   Owner

   Educator

   Minimum scope

   Only the exact approved objective text and codes needed for this assignment.

   Quality check

   A code must be present in the supplied authoritative document rather than inferred from memory.

Access and setup — proposed or unknown

Proposed minimum

1. Work in a new draft folder using a non-personal assignment copy and an output name that cannot replace the approved rubric.

2. Use an ordinary spreadsheet or CSV reader for arithmetic and schema inspection.

Compatibility unknowns

1. Unknown: host extraction quality for scans, installed skill behavior and whether a target LMS accepts the produced schema.

External actions requiring approval

1. Uploading school-confidential materials, importing into an LMS, replacing an approved rubric or distributing to students.

Why use this

1. An attractive rubric can still reward the wrong thing. If an assignment asks for reasoning but the strongest descriptor mainly counts words, students may be steered toward length rather than the intended learning.

2. The useful outcome is a transparent instrument that another educator can explain and challenge. It should expose unresolved choices before anyone's work is assessed.

Why an agent / simpler alternative

Useful judgment

Interpret the assignment's verbs and separate overlapping traits while keeping every proposed criterion tied to an explicit requirement.

Simpler alternative

Reuse a previously approved rubric when the assignment and learning target have not changed; check its arithmetic in a spreadsheet.

Unsuitable when

The task requires grading learner work, selecting accommodations, determining progression or making other consequential student decisions.

Expected deliverable and review checks

Proposed artifact

A blank rubric plus a decision sheet recording requirement links, intended total, actual total and unresolved wording.

Critical checks

1. Every criterion is supported by the assignment; presentation polish does not silently become a separate learning requirement.

2. Levels within one criterion describe the same observable trait and do not overlap in a way that makes two levels equally applicable.

3. Maximum points sum exactly to the approved total, weights reconcile, and machine-readable rows agree with the visible version.

4. A second educator can identify what evidence would distinguish adjacent levels without seeing any actual learner record.

Missing output evidence

No educator moderation exercise, completed rubric or LMS import was inspected.

First check — proposed, unexecuted

Status

proposed_unexecuted

Synthetic fixture

An original fictional adult workshop assignment asks for a short comparison of two supplied designs. The requested maximum is 10 points across three equally weighted criteria, and an obsolete attachment says 12.

Acceptance checks

1. The conflicting maximum is surfaced before a usable rubric is issued.

2. After a reviewer chooses 10, the allocation explicitly handles the remainder and sums to 10; no hidden rounding.

3. A presentation criterion is excluded unless the approved assignment actually asks for it; no learner work is requested.

Stop conditions

1. The educator has not resolved the policy conflict or real student work is introduced.

Failure signals and proposed responses

1. Signal

   The rubric's arithmetic differs from the declared maximum.

   Why it matters

   An apparently small discrepancy can change the meaning of every reported score.

   Proposed response

   Block use, recompute the distribution independently and show the educator the corrected totals.

2. Signal

   One requirement appears under several criteria.

   Why it matters

   The same feature could be rewarded or penalized more than once.

   Proposed response

   Map each requirement to its intended criterion and ask the educator to approve any deliberate overlap.

3. Signal

   A descriptor uses effort, attitude or an inferred personal quality.

   Why it matters

   The instrument would judge information that the assignment does not reliably reveal.

   Proposed response

   Replace it with observable work evidence or remove it for educator review.

Effort and cost — unknown

Setup work

1. Clarify the assignment version and scoring policy; remove student information.

Ongoing review

1. Moderate the blank instrument with another educator before each materially changed assignment.

Unknowns

1. Unknown: model cost, extraction effort, educator review time and inter-rater consistency.

Limits

1. This is a preparation aid. Teacher moderation is still needed; the first check would not establish fair grading or legal compliance.