How assessments are built and scored

8 reporting domains are scored. Coverage and difficulty vary by assessment.

• Analytical reasoning — Logic, evidence, assumptions and structured problem-solving.
• Quantitative and data literacy — Probability, estimation, financial reasoning and data interpretation.
• Communication — Comprehension, concise writing and audience-aware communication.
• AI and digital literacy — AI evaluation, privacy, source validation and responsible tool use.
• Cultural adaptability — Cross-cultural interpretation, adaptation and conflict resolution.
• Ethical judgement — Integrity, fairness, responsible data and technology use.
• Learning agility — Feedback, error correction and transfer to unfamiliar problems.
• Collaborative decision-making — Prioritisation, delegation and stakeholder decisions under uncertainty.

Once a completed attempt uses an assessment, test form or scoring model, that version is kept. Corrections create a new version instead of altering historical results.

Two pathways include a short written response: Postgraduate Readiness (one) and Professional Leadership (two). Each written item has a fixed rubric written with the item. An automated assessor scores the response against exactly those rubric criteria, and an independent second model checks borderline or low-confidence scores; the models may not invent criteria or change the grade boundaries.

A written item counts like any other scored item: its rubric score (from 0 to 1) is that item’s mark inside its competency, and the published competency weights then apply. If automated scoring cannot complete — the models disagree materially, the output fails validation, or the service is unavailable — the result stays pending and no certificate is issued until scoring completes; a score is never guessed. Once a result is issued, the candidate can appeal it through the appeals process, which can order a fresh review.

AI can draft items for review and help identify possible duplication. Drafts pass automated quality checks and recorded review before they enter the active bank. Scoring uses pinned rules. Items can later be revised or retired without changing historical results. Model, prompt and reviewer records are versioned.