Authentic assessment — a crash course

Decide what students should be able to do, then design the task that makes them do it. Misalignment — teaching one thing, assessing another — is the most common assessment fault.

• Write the task from the learning outcome backwards: what evidence would show the outcome is met?
• Align teaching, activities and assessment to the same verbs (analyse, design, evaluate).
• Cut assessment that doesn't evidence an outcome — less, but better.

Authentic tasks mirror what graduates do in practice — they're more motivating, and far harder to outsource than a generic essay or quiz.

• Give the task a real audience, context, or artefact (a brief, a case, a portfolio, a plan).
• Assess the process as well as the product — drafts, reflections, decisions.
• Build in professional judgement, not just recall.

Generative AI raises the floor on what can be produced unaided. The durable response is task design, not detection — assess what AI can't do alone.

• Foreground process, iteration, in-class or oral components, and personal context.
• Be explicit about where and how AI use is permitted, and require disclosure.
• Use NTLSN's GenAI assignment redesign tool to stress-test a task for AI-vulnerability.

No single task carries the load. Mapped across a program, assessment can reduce over-assessment, build skills cumulatively, and evidence the degree's outcomes.

• Map where each outcome is introduced, developed and assured across the program.
• Reduce duplicated, low-value tasks; sequence toward a capstone.
• Moderate across markers so a grade means the same thing.