What is AI-Safe?

A non-exam assessment can be designed with the following in mind:

Our concepts	Is your assessment…	What we mean
	context-specific?	The context is clearly defined and relatively limited in its scope.
	authentic?	The work done is similar to what professionals often need to do.
	collaborative?	There is a joint effort requiring equal commitment from everyone.
	*process-driven*?	How work is done is as important as its end-product.
	*generative*?	New information, ideas, artefacts or designs are produced.

To make their assessment as AI safe as possible, a designer should be able to answer “yes” to (almost) all of the above.

Whether it is for an assessment task or a problem-based activity, the design will include multiple layers of protection against AI.

How AI-Safe is structured

There are five components, each one exploring your assessment design in relation to a single concept in our framework.

This is done through the key questions, which are supported by a series of further guiding questions.

In order to provide the fullest protection against AI misuse, an assessment designer should consider these questions when designing a task or a problem-based activity (click to expand):

Is the assessment context-specific?

– Is it set in a specific context?
– Does it have specific objectives?
– Is it clearly limited to a specific field of (professional) activity?
– Can its focus be more local than national or international?
– Would learners have an active role in this local context?
– Could any other aspect be made more specific?

Is it authentic?

– Does it involve solving (a) problem(s) for a practical / professional purpose?
– Do learners have to make decisions to solve the problem(s)?
– Do they act from the standpoint of people who normally make such decisions?
– Do they need to apply relevant knowledge and skills?
– Do they need to have some initiative and judgement?
– Are they required to justify their decisions?

Is it (partly) collaborative?

– Do learners work together (whether for part or the whole of the assessment)?
– Are key decisions made collectively in a group?
– Is the choice of roles, responsibilities, personal objectives also made collectively?
– Is there a system for monitoring / recording individual contributions to the group?
– Do learners participate in a review of their individual performance / task achievement?
– Is each learner required to give evidence of their own involvement / achievement(s)?

Is it process-driven?

– Do learners need to produce e.g. text, calculations, code, designs, images, speech, etc.?
– Can the production process be broken down into steps or stages?
– Can (at least some of) these be directly observed or recorded and documented?
– Is (at least part of) the assessment focused more on its process than its final product(s)?
– Does this process involve bringing different elements into a coherent whole?
– Could it be linked to other processes or products in the same course?

Is it generative?

– Is it built on information / ideas emerging from work that learners do for themselves?
– Is such content produced in the context of the assessment itself?
– Do learners make decisions on the basis of the content they have generated?
– Is the learner-generated content used in different parts of the assessment?
– Can the use made of this content be observed through the assessment process?
– Do learners account for how their information / ideas have been generated?

To find out about the terminology in these questions (and the thinking behind our conceptual framework), see here.

The overlap with principles and propositions from national education bodies has been summarised here.

Examples of assessment ideas that may be applied to various disciplines can be found here, on page 2.