AI-proofing assignments

If the info isn’t online, AI can’t use it to generate content:

• Contextualise the task: Link the task to a specific context that AI can’t generate
  • Link the assignment to personal experience(s). For example, reflect on a clinical case study from your internship; write a research proposal on the impact of urban lighting on insects in your own neighbourhood.
  • Base assignments on your own teaching practice or learning environment. Focus the assignment on discussions or conversations held during a lesson or on material presented during the lesson. For example, analyse the debate that took place during the lesson on, say, 15 March.

• Work with sources which are secure or local. Allow students to use materials that are not publicly available online, such as internal documents. For example: analyse the University of Antwerp's inclusion policy using the internal policy document.
• Use self-collected or context-specific data. Encourage students to collect their own data or work with customised data. For example: interview three students from different study programmes about their experiences with X; collect measurement data in a practical and analyse them, keeping in mind measurement errors and the influence of the local context (e.g. temperature in the classroom).
• Work with non-searchable assignments. Give tasks that AI finds difficult to reproduce, such as fictitious cases, future scenarios or self-created prototypes.

Most AI systems have limited or delayed access to current data. If necessary, ask explicitly how far back their data goes (this varies by tool and is constantly evolving). Moreover, some AI tools also claim they can perform live searches to retrieve recent information (such as ChatGPT on 30 April 2025: ‘I can run live searches to give you an up-to-date answer’).

• Choose formats that are less accessible to AI, such as videos, academic posters, models or other multimodal products. Please note: AI tools are getting better at handling the variety of formats.
• Use live and interactive forms of assessment such as discussions, debates, oral reasoning or performance tasks. These require presence and direct interaction, which limits AI assistance.
• Provide variety: combine different forms of assessment within the course unit (e.g. quizzes, short essays, presentations). AI performs well at specific tasks, but struggles with varying formats and expectations.

Such as problem-solving, creativity, critical thinking, reflection, analysis, interpretation, synthesis and assessment. The simpler the task (e.g. summarising, describing or explaining), the more likely AI is to generate strong output.

These have a contextual, personal character and rely on unique, personal insights of the student. Here, the teacher acts as a coach and feedback giver, with a clear view of the student's learning process. This personalised approach makes it particularly difficult for AI to offer meaningful support.

o    Request interim versions or organise an interim survey. Ongoing assessment makes the learning process more trackable and reduces the likelihood of students relying solely on AI. They have to then continually demonstrate their grasp of the subject matter.

o    Provide an intermediate and final presentation. This way, you get an idea of their progress and ownership throughout the process.

o    Give interim feedback and let students:

• reflect on how they processed this feedback,
• justify why they did or did not follow certain suggestions.

Ask students to describe their approach: what steps they took, what choices they made, what obstacles they encountered and how they overcame them. Keeping a logbook works well here.

Also evaluate the process, not just the product. So focus not only on the end result, but also on the (learning) process.

Use an oral explanation to elaborate on the competences to be tested and necessary insights. In doing so, delve into the process if necessary: how did students arrive at their results? What were their considerations in developing, writing or resolving? What sources did they use, and why? How do they justify their choices, approach or solution?

Have students complete assignments during lessons or in a controlled environment without access to AI. Consider, for example, writing an essay or programming code in a room without internet. Consider integrating this into a learning trajectory where students first learn fundamental skills, such as academic writing or programming, in a protected context. At a later stage of the learning process, the use of AI can then be purposefully introduced.