Assessment that earns its place in the AI era
- Naomi Rowan
- 4 days ago
- 4 min read
Updated: 1 day ago
AI has made some assessment tasks feel more vulnerable, but that doesn't mean every traditional essay or exam is suddenly obsolete. Essays and exams can still hold a valuable place when testing fluency, independent argument, disciplinary knowledge, synthesis, recall and performance under pressure.
At the same time, some assessment tasks now need to justify their purpose more clearly. If a task can be completed convincingly by a general-purpose AI tool with little evidence of student judgement, process or contextual understanding, the institution has a design question as much as an integrity question.
The key question becomes: 'What is this assessment really trying to evidence?'
Knowledge still matters. If anything, students need enough disciplinary understanding to question what AI produces. In a world of abundant, searchable knowledge, students need to know what is useful, what is trustworthy, and how to use it well.
That requires judgement, critical thinking, professional reasoning and ethical decision-making. It also requires metacognitive awareness: the ability to monitor their own understanding, recognise uncertainty, respond to feedback, adapt their approach and explain the choices they have made.
Application, creativity, collaboration, communication and the ability to work with uncertainty therefore become increasingly important, not as generic “skills”, but as part of how students show what they understand and what they can do with that understanding.
These skills need different kinds of evidence.
A final essay may still be right for some purposes. A secure exam may still be necessary in some contexts. A project, portfolio, oral explanation, staged task, design rationale, case analysis, practical output or reflective commentary may work better elsewhere.
The need is to understand what the assessment is trying to evidence and whether the design still supports that purpose.
This is where authentic assessment becomes useful, though the term is sometimes used too loosely. It does not simply mean “make it like the workplace” or “set a project.” Meaningful authentic assessment asks students to use and apply knowledge from their course, make their thinking and decisions visible, and demonstrate the wider capabilities required for the complexity of the discipline or profession - while still meeting academic standards.
For example:
A project can be rigorous if the criteria are clear, the process is staged, and the student’s decisions and thinking are visible.
A portfolio can be rigorous if it involves selection, reflection, justification and development over time.
An oral component can support verification if it is structured, moderated and accessible.
A group task can be meaningful if individual contribution, collaboration and judgement are made visible.
A traditional exam can remain appropriate if it is the right fit for the learning being assessed.
AI-era assessment design needs this kind of nuance.
A purely defensive response fuelled by fear of cheating or unmanaged student AI use can push institutions towards more restriction, surveillance and high-stakes assessment without necessarily improving learning. In some cases, it may even narrow the learning experience.
A more constructive response builds metacognition into the assessment. It asks students to make their thinking visible, demonstrate progress through feedback, explain choices, monitor their own process and reflect on how their understanding has changed.
Universities still need rigorous assessment of learning, after all, standards, progression and awards depend on defensible judgement. They also need to prioritise assessment for learning where students are given opportunities to practise, receive feedback, revise, explain, apply and improve. The strongest assessment strategies usually need both.
That might mean building in a proposal, annotated bibliography, feedback response, process log, oral explanation, data commentary or reflective judgement. In other contexts, the task might shift towards a prototype, case response, applied decision or practical output.
Assessment shouldn't become complicated or unwieldy. It can, however, be reworked to create better evidence of learning and more active student engagement.
AI can still be used within authentic or staged assessment, so this isn't a magic shield against misuse, but it changes the nature of the conversation. Instead of asking if AI was used, the assessment can ask how: what the student understood, how they challenged the output, how they made decisions, how they used tools critically, how they responded to feedback, and what they did with the knowledge in context.
That is a more educationally useful conversation, and it develops the kind of judgement students will need beyond the assessment itself... It also needs operational support.
A well-designed assessment can still fail if the workflow around it is weak. If students don't understand expectations, the task becomes confusing. If staff don't have time to give useful feedback, assessment for learning becomes aspiration rather than practice. If Moodle or another platform doesn't support the stages, the process becomes manually heavy. If moderation is unclear, confidence in standards suffers.
Assessment design and assessment workflow need to be thought about together.
AI should not push universities into panic redesign. It should create space for more honest questions about what assessment is for, what evidence is needed, and how students can be asked to show judgement in ways that are fair, rigorous and workable.
The goal isn't novelty or restriction, but meaningful assessment that earns its place.
Comments