Concept GenAI and assessment in higher education
In higher education, many assessments take place without supervision. Even before GenAI, this raised questions about validity. Now, with over 70% of UvA students using GenAI at least weekly, these concerns are urgent. Studies show that frequent GenAI use can reduce critical thinking through cognitive offloading. Because assessment shapes how students learn, design choices matter more than ever.
Designing assessments that maintain integrity and foster learning
Generative AI is (becoming) part of academic and professional practice. Making assessment completely ‘AI‑proof’ is neither realistic nor desirable. The key is to safeguard educational quality while deciding when students may – and may not – use GenAI.
Designing assessment for the AI era means revisiting the constructive alignment in your course: which learning outcomes must students demonstrate without GenAI, which involve responsible use of it, and what balance between supervised and unsupervised assessment best fits your discipline and level?
The two-lane approach
To address this question, the TLC recommends the two-lane approach (Liu & Bridgeman, 2023; Npuls, 2025). This model distinguishes between learning outcomes that require independent mastery and learning outcomes where students must learn to use GenAI tools critically and responsibly. This approach helps ensure that assessment remains both credible and relevant to modern professional practice, where responsible AI use is increasingly expected.
For learning outcomes that students master independently, without using GenAI
Use when:
- the learning objective requires independent reasoning or skill
- the task could easily be outsourced to AI
- the assessment must verify procedural or practical competence.
Learning outcomes where GenAI use is allowed or part of the learning process.
Use when:
- AI use reflects professional or disciplinary practice
- the learning outcome focuses on evaluating, refining, or integrating AI-generated output
- the learning objective can still be achieved even if AI tools are used
- the assignment aims to develop AI literacy, such as critical evaluation of AI outputs or responsible use of AI tools.
How Lane 1 and Lane 2 work together in practice
Determining which lane applies depends primarily on the learning goals of the course, and how learning activities and assessment support those goals.
Lane 2 assessments (unsupervised) are valuable, but cannot on their own provide sufficient evidence of learning. Because students may use AI, the final product does not reliably show what they can do independently. For this reason, any learning that must be demonstrated without support needs to be verified in Lane 1 (supervised).
The key design question is therefore not whether to use unsupervised assignments, but: What do I want students to demonstrate independently—and where will I verify that?
This checklist helps you find out whether your course assessment is vulnerable to misuse of AI.
GenAI and academic misconduct
If you suspect that a student has used GenAI inappropriately in an assignment or exam, please inform the exam board. They can investigate the matter and take appropriate action, helping to ensure equal treatment of students.
Do not rely on AI detection tools. In their current form they are not reliable and cannot accurately determine whether a text was written by a human or generated by AI. In addition, they may disadvantage students writing in a non-native language. For these reasons, their use is generally discouraged.
Coordinating assessments at programme level
Designing strong individual assignments is only part of the solution. Because many programmes use multiple written assignments in different courses, it is important to look at how assessment works across the curriculum. A programme-level view helps create a balanced curriculum and reduces over-reliance on any single assessment format.
Programme-level coordination starts from the intended learning outcomes. Mapping learning outcomes and assessment formats per course helps you see:
- where independent mastery is assessed (Lane 1)
- where students develop AI skills and use AI responsibly (Lane 2).
The two-lane approach works best when assessment is coordinated across courses. In programmes with many unsupervised written assignments, programme teams can review:
- where written assignments are most appropriate
- where alternative formats better demonstrate independent learning
- where similar assignments appear and could be streamlined.
Programme teams then decide together how each course contributes to the overall assessment strategy. Earlier courses may focus more on demonstrating foundational knowledge independently, while later courses increasingly integrate AI-supported research or professional practice.
Advice on adjusting your assessment, contact us!
If you would like to discuss potential adjustments to your assessment design, you can contact your faculty’s assessment specialists via TLC Contact. You may also seek advice from the assessment specialists at TLC Central.
When making adjustments to assessment design, it is advisable to inform your programme director, so that potential risks and changes can be considered at programme level.