Muhammad Tayyab Ilyas

Chapter 6. Fair Assessment in the AI Era

In a senior physical education course in Toronto, students used AI to help organise a warm-up plan. Their teacher then required something the plan alone could not supply. Students had to ‘lead and explain the routine in person’ (C21 Canada, 2025, p. 5). The written work remained, alongside a performance and an explanation.

This school account comes from a report on teachers exploring an AI Assessment Scale. Leon Furze, one of its developers, helped train them. The report describes practice and teachers’ reflections, without a controlled test of learning. Some redesigned tasks never reached full implementation, and individual coaching took considerable time (C21 Canada, 2025, pp. 4–5).

Still, the physical education example makes the assessment question unusually clear. A student could arrive with a polished plan and then struggle to explain or lead it. Equally, someone might use help with planning and still demonstrate the intended skill. The teacher needed evidence beyond the document to distinguish those possibilities.

What the chapter covers

  • What the mark allows us to claim
  • Why a detector cannot settle it
  • Change the task and its conditions
  • Drafts, checkpoints and short oral checks
  • Help with marking, care with judgement
  • In practice
  • When resources are thin
  • Where this chapter leaves us
  • Questions for reflection
  • Framework connections

Questions for reflection

Take these questions to a phase or subject meeting, with one recent assessment in front of you.

  1. What claim about understanding does its mark allow you to make, and which assumption about help supports that claim?
  2. Which brief classroom demonstration could strengthen the evidence without consuming another lesson?
  3. How would your response differ between uncertain understanding and proven misconduct?
  4. Could a student using an agreed access arrangement demonstrate the same skill through your proposed check?
  5. If you tried AI marking, what evidence would show whether its suggestions improved your decisions about learning?