Our concepts	Our reasons	To consider
context	The more the context is specific and limited, the less chance there is that AI can complete tasks or solve problems on learners’ behalf. Since AI lacks conceptual understanding, its abilities will rest on the material that is available online.	Identify suitable contexts about which little or no data, information or commentary is accessible to AI. The scope of such contexts will typically be small-scale and/or local (as well as appropriate to the discipline(s) concerned). 1
authenticity	The exposition of factual content and ideas is something that AI can easily produce. Topic-based or theoretical assessments are highly vulnerable as a result, especially in areas of general interest.	Focus more on solving real-life problems in a practical manner than on displaying knowledge and skills in a way that AI can imitate. By assessing the application (and not the display) of knowledge and skills, an extra layer of protection against AI misuse can be added. 2
collaboration	Learners may appear to produce or present their own work for assessment purposes when in fact they are regurgitating AI-generated material. The need for collaboration imposes constraints on what learners include in their assessed work. 3	Introduce group decisions and accountability so as to hinder learners from passing off AI-generated work as their own. Different patterns of collaboration can be built into the design of an assessment, in accordance with recognised practice in its vocational field. 4
process	Learners’ agency and ownership comes about as a result of their engagement with the process through which their work is created. Once this process becomes an integral part of their assessment, it can be monitored, tracked and/or observed so that individual learners’ work can be fully identified and assessed.	Integrate an appropriate and fully defined process into the assessment design so that any use of AI by learners may be transparent and open to scrutiny. This process can also be geared to the assessment of individual learners’ performance. 5
generativeness	AI not only eliminates the need for learners to produce material for assessment, but also the need to enquire, experiment and reflect in order to identify suitable content for this material.	Include the use of information, ideas etc generated through the learners’ own effort instead of relying on AI as a substitute for independent thinking. The process on which the assessment is based will enable this emergent content to be included in the final product as evidence of learning. 6

1. It will be possible in many disciplines to address global, international, or national issues through a task or problem-based activity set in a local context. Alternatively, the scope of an assessment may be limited in other ways, e.g. by viewing an international scenario from a local perspective. ↩︎
2. It should be possible to determine to what extent an assessment design resembles the ways that knowledge and skills are used in the real world. ↩︎
3. With the right tools and criteria, individual performance will be assessable in a collective context. ↩︎
4. A task or problem-based activity need not be 100% collaborative. For example, an assessment may start off as collaborative work in the planning stages and continue through individual work within a negotiated group plan. ↩︎
5. A well thought-out process can reveal exactly how individual students have worked on problem-solving and meeting task objectives. ↩︎
6. AI can be prompted to generate material with embedded information that may be relevant to a particular assessment. Yet if all that is required of learners is that they submit such material, there will be no evidence of their ability to process and make use of the information it contains. ↩︎

Pages: 1 2