Engineering Assessment Levels: Making Score Differences Understandable

By Ahmad Issam Maadarani

A zero-to-ten form does not automatically make its numbers clear. This article proposes levels described through observable performance, then testing whether users distinguish and apply them to new cases. It proposes scale design rather than universal quality numbers or proof that more categories improve precision.

1. A number without description leaves meaning open

One judge's 7 may mean requirements met with limited weakness; another's may mean substantial improvement needed. Numerical disagreement then concerns scale meaning before dish quality. Define success and deficiency within the task before selecting distinguishable levels. Where two levels have no explainable difference, review the distinction rather than retain it because it is familiar.

2. Four elements of a level description

Specify required function, observable indication, extent and experiential effect. For cohesive bites, one level may describe preserved structure during sampling, another fragmentation obstructing the function. Criteria need not all share one description method. Do not smuggle undeclared properties into general quality: levels explain criteria rather than expand the task.

3. References are not dishes to imitate

Training cases can illustrate a level without making their colour, shape or style hidden requirements. The reference illustrates one property, not the only good dish. Use varied styles achieving the same function and borderline cases revealing ambiguity. Explain what each example represents and excludes. Choosing the most visually similar dish demonstrates image memory rather than criterion application.

4. Hypothetical case: three cohesion levels

Consider three training descriptions without final scores. First: the required bite can be taken without obstructive disintegration. Second: local fragmentation appears, but most required bites remain accessible through the declared method. Third: fragmentation prevents the required bite at examined locations. These illustrate rather than establish universal food grades. Review task fit and fairness across presentations. If a component is intended to be eaten fragmented, this scale is unsuitable rather than evidence of poor performance.

5. Boundaries and insufficient information

A sample may lie between levels or provide insufficient information. Predefine procedures: nearest level with reasons, permitted re-examination or non-assessable status. Do not invent decimal places to conceal ambiguity or automatically equate missing samples with lowest performance. Separate failure to provide a required item from inability to observe it. Procedures depend on task and rules rather than a universal competition policy.

6. Test with cases absent from development

Users independently record descriptions, then review agreement and disagreement. Matching numbers alone are insufficient: examine whether the same property supports them. Frequent boundary disagreement calls for reviewing wording, examples or sampling. New cases avoid testing memorisation. Limited testing cannot validate the scale for every product; record scope, cases and untested areas, preserving the version used.

7. Version control preserves the task

Give the scale a version, date and change record. Do not transfer old results to a new scale as if meanings were unchanged. Competition ambiguity follows declared procedures applying to all affected entries rather than one favoured dish. Separate future system improvement from correcting application of the current rule. Descriptions matter as much as numbers because they define participant and judge expectations.

8. Conclusion: fewer levels may be clearer

Scale quality rests on meaningful distinctions, not length. When judges explain a placement and cooks understand what advances performance, assessment supports development. An unexplained number may display precision without producing reviewable knowledge.

9. A concise working sheet

Suggested citation: Maadarani, Ahmad Issam (2026). “Engineering Assessment Levels: Making Score Differences Understandable” Methodological article, version 1.0.

© 2026 Ahmad Issam Maadarani. All rights reserved, subject to applicable legal exceptions.