What concepts have we excluded?

AI-Safe is entirely consistent with standard principles of assessment, such as validity and reliability. But as with these principles, there may be some tension and possibly a trade-off between concepts when our framework is applied to specific assessments. The aim is to achieve an acceptable balance overall.

There are also well-known concepts of assessment design which do not feature explicitly in our framework:

• critical thinking
• reflection
• creativity
• innovation

AI is already capable of generating material that may appear to be critical and reflective, or creative and innovative. While its ability to produce human-like content is purely imitative and not based on conceptual understanding, we cannot rule out the possibility that such content might be good enough to meet assessment criteria for creativity or criticality.

It is in fact the contextualisation of what learners are required to do that will determine to what extent this type of content is vulnerable to AI misuse. Likewise, the other concepts in our framework provide additional safeguards for the legitimacy of assessment work.

Pages: 1 2