SEA-EU · Task 2.2 · Assessment and generative AI
Open by Design
Assessment in the age of generative AI, and how to keep open exams worth having.
A student with a current AI tool can hand in an answer that most examiners can’t tell from a good student’s. Even so, this course doesn’t argue for moving everything back into the exam hall. It asks how you can design take-home exams, assignments and projects so that they still tell you what a student has learned. Those formats give students time to go deep in their reading and writing, and they are where students can learn to use AI well.
§01 The course in brief
Four ideas carry the course
No right answer
Some answers to AI in assessment are better and some are worse, but there is no single solution. Each one trades something away for what it gains.
Structure, not rules
Rules about AI use can’t be enforced where nobody is watching, and detection doesn’t work. So the line has to be drawn between kinds of assessment, since it won’t hold inside a single one.
The open lane is worth keeping
Take-home exams, assignments and projects give time for depth. They are also where students can learn to work with AI in their subject, critically, while they learn the subject itself.
Validity comes from the whole
It builds up across the programme through process, dialogue and secured points. Trying to control every single exam won’t produce it.
§02 How it works
One assessment, seven steps
You bring one assessment you know well, in a course you teach or are responsible for. Each step ends with a short worksheet about it, and the panel Your assessment collects what you write. At the end you have a one-page brief for one change you could try.
What you write stays in your browser. Nobody else sees it until you copy it somewhere yourself.
- StartBring one assessment15 minChoose your assessment, then mark four exam answers and find out how each student worked.
- Step 1Why the old question fails25 minWhat AI can do today and why rules and detectors can’t carry validity. Then the two lanes.
- Step 2What the open lane is for20 minHybrid practice: learning the subject and learning to work with AI critically, at the same time.
- Step 3Studio: make the process visible25 minWindows into the black box, real problems, and an AI answer to fix.
- Step 4Studio: talk with the student30 minA short conversation, five good questions, and assessment twins.
- Step 5Beyond your course25 minHoles that shouldn’t line up across the programme, two national answers, and what happens to the grade.
- Step 6Stress test and commit20 minNo right answer and three permissions, then one change you can try.
§03 Go deeper
Six closer readings
Each step points to a longer page on one of its sources. They are optional, and you can read them in any order.
- Step 2From cognitive necessity to cognitive choice15 minMontebello’s engagement gap and five principles for assessment that puts thinking at the centre.
- Step 3Black Box Assessment15 minMaking the learning process part of the evidence.
- Step 4Assessment twins15 minPairing an open task with a second assessment of the same outcome.
- Step 5Triangulation15 minCombining kinds of evidence into one judgement.
- Step 5Control or learning?15 minNorway’s committee and TEQSA, and what each answer protects and puts at risk.
- Step 5What AI means for the grade15 minWhat a grade can still tell you after AI.
§04 Your brief
What you have so far
The brief grows as you work through the steps. Copy it into a document, download it, or print it.