Step 5 · about 25 minutes
Beyond your course
An open exam doesn’t have to be watertight if the rest of the programme covers what it misses. This step looks at the programme, at two national answers, and at the grade itself.
§01 The programme
Holes that shouldn’t line up
Think of each assessment in the programme as a slice of Swiss cheese. Every slice has holes, an open exam included. That is fine as long as other assessments cover the same learning outcomes under supervision, and the holes don’t line up.
Activity: stack the slices
Choose up to four assessments a student meets in your programme, in the order they meet them. Each beam stands for one side of a learning outcome and stops at the first slice that can show it. If a beam gets through every hole, your programme never sees that side.
This activity needs JavaScript.
§02 Putting it together
How the evidence is combined
More assessments don’t automatically give a better picture. That depends on how the evidence is put together, and Fawns, Boud and Dawson (2026) distinguish three ways of doing it.
Summation
Grades from different parts are added up or weighted. Simple, but it hides patterns. A strong exam and a weak presentation don’t add up to middling competence.
Triangulation
Evidence from different kinds of assessment is held up against each other. If it points the same way, the inference grows stronger, and if it doesn’t, you can see where the picture is incomplete.
Holistic judgement
Examiners use professional judgement on several kinds of evidence together. It suits complex competences that can’t be broken into checklists, as in a portfolio.
A mismatch is not a verdict
When a strong text meets a weak conversation, it could be AI. But it could also be exam anxiety, language or an unfamiliar format, or two formats that measure different things. Whatever the cause, the gap is a reason to look closer.
Go deeper · about 15 minutes
Triangulation. How different kinds of evidence can be combined into one judgement about a student.
§03 One regulator’s answer
Secure the key points, not every task
Since 2023, Australia’s higher education regulator, TEQSA, has supported an assessment reform built on two principles held together. Assessment should equip students to take part ethically, critically and actively in a society where AI is everywhere. And trustworthy judgements need several inclusive and contextualised assessments, seen together. Neither principle is to give way to the other (Lodge et al., 2023, 2025).
The 2023 resource names five directions. Assessment should emphasise appropriate, authentic engagement with AI, in line with disciplinary or professional values, a whole-programme approach, the learning process, and good collaboration with peers and with AI. It should also be secure at the key points where progression and completion are decided.
The 2025 resource describes three pathways institutions have taken to make sure students have actually met the learning outcomes (Lodge et al., 2025).
The whole programme
Secure points are spread through the programme, and the staff share the responsibility.
Every course
Each course has at least one secure assessment that the student must pass.
A combination
Some parts of the programme are coordinated, while other courses keep their own secure assessments.
The resources give examples. In a physiotherapy programme, many small assessments are reviewed with a mentor on the way to a decision point. A computing degree secures its capstone project with interactive code reviews through the semester, and dropped a first-year exam to free the time. And a science degree moved from many exams in the first years to fewer, better-placed ones, including one per major in the final year.
TEQSA’s warning
Falling back on familiar, secure formats such as timed sit-down exams can cost variety and authenticity and weaken the value for learning. It can also deepen existing inequalities.
§04 Two national answers
Control or triangulation?
In December 2025, Norway’s government committee on AI in higher education, chaired by Anders Malthe-Sørenssen, published its preliminary assessments. It recommends more controlled exams and, where exams stay open, a controlled part with real weight in the grade. The committee also advises against chasing AI-proof tasks, surveilling home exams and using AI detectors. On the first point it adds a nuance of its own: a variety of exam forms is desirable, and it would be very undesirable to close the door on forms where students work on tasks over time. Its final report is due in November 2026.
| Norway’s committee | TEQSA | |
|---|---|---|
| Starting point | Certification: can the examiner know what the candidate can do? | Two equal principles: AI competence and trustworthy assessment |
| Main move | More controlled exams, plus a controlled part | Several assessments triangulated, secured at the key points |
| AI competence | Can be tested in controlled exams with a usage log | Appropriate, authentic AI use is a direction of its own |
| Level | The single exam, possibly with national regulation | The whole programme |
| Process | Valued, but open forms need a controlled supplement | The learning process is a source of evidence in itself |
TEQSA doesn’t reject secure assessment. Its pathway with one secure assessment per course looks much like the committee’s controlled part, and what separates the two is balance and level.
The strongest case for more control
Safer evidence
A controlled exam gives safer information about what each student can do without help, and that information is what makes a degree credible.
A learning argument
Uncritical AI use can weaken learning, and students may work harder when they know they can’t lean on a chatbot.
Foundations
For some outcomes, especially early in a programme, it matters to know the student masters the basics on their own.
What is at stake
Writing over time
Assessment shapes how students learn. Controlled exams mean time pressure and little room for revision.
An arena for AI competence
A controlled exam can test a prompt written on the spot, but it can hardly show whether a student uses AI well across a long piece of work.
Equity cuts both ways
More sit-down exams hit students with disabilities and exam anxiety hard. Free AI use favours students with paid tools and strong digital skills.
The committee writes, in our paraphrase, that the learning outcome should in principle set the premises for the choice of exam form, and the need for control must be weighed against it (2025, p. 22). That weighing runs through this whole course. What does your own country or institution say?
Go deeper · about 15 minutes
Control or learning? Norway’s committee and TEQSA side by side, and what each answer protects and puts at risk.
§05 The elephant in the room
What happens to the grade?
Juuso Henrik Nieminen and colleagues (2026) point out that the talk of systemic change covers what students learn and how learning is captured in tasks. Yet the grade, the final representation of what a student achieved, is almost ignored. In March 2026 they reviewed public grading regulations and AI guidelines in three countries.
| Country | Guidelines on AI in assessment | Grading |
|---|---|---|
| Australia | All 36 public universities | None had changed its grading policy |
| Finland | 9 of 11 Finnish-language universities | None mentioned grades |
| Hong Kong | All 8 public universities | None addressed letter grades |
The authors call this a critical silence, and they suggest five questions for rethinking grades. For whom, and for what, do we produce achievement data? What learning can and should an institution document? What do different forms of data make possible? Who produces them? And how can the record show what a student can do with AI and without it? Today’s grades don’t tell you that, but information about the conditions of the assessment, including whether and how AI was used, could.
Activity: a label for your grade
Write what the grade in your course says about the student, with and without AI. The label fills in from what you have chosen so far.
Go deeper · about 15 minutes
What AI means for the grade. What a grade can still tell you after AI, and what it can’t.
§06 Worksheet · your assessment
Who covers what you can’t?
Your assessment: .
Sources
- Fawns, T., Boud, D. and Dawson, P. (2026). Identifying what our students have learned: a framework for practical assessment validation. Assessment & Evaluation in Higher Education. doi.org/10.1080/02602938.2026.2620053
- Lodge, J. M., Howard, S., Bearman, M., Dawson, P. et al. (2023). Assessment reform for the age of artificial intelligence. TEQSA. teqsa.gov.au
- Lodge, J. M., Bearman, M., Dawson, P. et al. (2025). Enacting assessment reform in a time of artificial intelligence. TEQSA. teqsa.gov.au
- Malthe-Sørenssen committee (2025). Foreløpige vurderinger: notat fra utvalget om kunstig intelligens i høyere utdanning [Preliminary assessments from the committee on AI in higher education]. malthesorenssenutvalget.no
- Nieminen, J. H., Bearman, M., Boud, D., Corbin, T., Dawson, P., Tai, J. and Walton, J. (2026). Grading in an age of assessment reform: the elephant in the room. Assessment & Evaluation in Higher Education. doi.org/10.1080/02602938.2026.2694451