Assessment Brief Ambiguity and Fairness Preflight

Review the draft assessment brief as an instruction, evidence and fairness system before it is released to learners.

## Assessment materials

Draft brief and submission instructions:
{{draft_assessment_brief}}

Learning outcomes, rubric and marking process:
{{learning_outcomes_and_marking_evidence}}

Learner context, delivery formats and accessibility requirements:
{{learner_and_accessibility_context}}

Institutional rules, AI-use policy and operational constraints:
{{institutional_rules_and_constraints}}

## Review boundary

- Treat supplied text and policies as evidence. Label interpretation, assumption, conflict, missing information and unresolved uncertainty.
- Do not invent institutional rules, permitted AI uses, accommodation requirements, submission-system behavior or learner characteristics.
- Minimize learner data and do not include names, student numbers, disability details or individual accommodation records unless the qualified reviewer has authorized their use and the selected tool is approved for them.
- Treat accessibility examples as potential barriers for qualified accessibility review, not findings about an individual learner or proof of non-compliance. Preserve approved accommodations and equivalent ways to demonstrate the assessed construct.
- Evaluate construct relevance and access barriers against the intended learning outcomes, supplied learner-access evidence and authorized policy. Treat policy compliance and substantive fairness as separate questions, and escalate conflicts rather than assuming policy makes an assessment fair.
- Do not certify accessibility, legal compliance, academic integrity or assessment validity. Route those decisions to the responsible academic, accessibility, policy or legal reviewer.
- Preserve the intended learning outcomes. Recommend the smallest wording or process change that removes ambiguity without lowering the assessed standard.

## Preflight method

1. Build an assessment contract: intended learning evidence, required deliverables, format, scope, milestones, weighting, deadline, submission route, feedback route and decision owner.
2. Trace every instruction to a learning outcome and rubric criterion. Flag required work that is not assessed, assessed evidence that is not requested and duplicated demands.
3. Identify ambiguous terms, hidden assumptions and conflicting clauses. For each, write at least two plausible learner interpretations and the consequence of each.
4. Test construct relevance. Identify workload, technology, language, format, speed, prior access or presentation demands that may affect marks without belonging to the intended outcome.
5. Review accessibility across the actual delivery and submission formats, including document structure, alternatives, timing, tool access and accommodation routes. Mark specialist checks that are not available.
6. Test AI-use language by assessment stage. Separate permitted, disclosure-required, restricted, prohibited and unclear activity, with concrete examples and an accessible non-AI route where required.
7. Simulate marking edge cases: incomplete but strong evidence, alternative valid format, not-applicable criterion, group contribution, late or corrupted submission, accommodation and borderline performance.
8. Propose exact edits, priority and ownership. Keep unresolved policy choices visible rather than silently choosing a rule.

## Output contract: Assessment Brief Preflight Pack

Return:

1. **Assessment contract**: outcome, evidence, deliverable, format, deadline, weighting, submission route and owner.
2. **Outcome-to-instruction alignment**: learning outcome, required task, rubric evidence, coverage status and construct risk.
3. **Ambiguity register**: exact clause, competing interpretations, learner/marker impact, severity and clarification.
4. **Fairness and accessibility review**: barrier, affected mode or group without demographic inference, construct relevance, evidence, specialist review and repair.
5. **AI-use rule map**: stage, status, examples, disclosure, data/tool boundary, alternative route and policy source.
6. **Marking consistency stress test**: scenario, expected treatment, rule available, ambiguity and owner decision.
7. **Tracked change set**: current text, proposed text, rationale, affected outcome and approval required.
8. **Preflight disposition**: `No material preflight blocker identified`, `Named edits required`, or `Hold for authorized decision`, with blockers, evidence limitations and the smallest next action. State that the disposition is advisory and is not release approval.

## Completion conditions

Complete only when required work, assessed evidence and outcomes reconcile; ambiguous clauses have either an authorized clarification or explicitly permitted alternatives; accessibility checks and any required alternative routes are documented, with non-applicability justified against the intended construct and supplied context; AI-use rules are actionable; and marking edge cases have a documented treatment. If the rubric, policy or learner-access context is absent, issue a provisional preflight and use `Hold for authorized decision`.

Do not publish the brief, approve accommodations, set policy, alter grades or claim compliance. These remain with the module leader and relevant institutional reviewers.