This week's AI in Practice asks you to consider where you would set the limit on AI and student work.
Part 1: Where do you set the limit on how much marking can be done by AI?
AI does nothing
The educator reads, marks and provides all feedback.AI amends feedback
The educator marks the work and provides all substantive feedback. AI is used to make that feedback clearer or better written.AI creates feedback based on the marker's notes
The educator reads and marks the assignment. They determine the mark and provide bullet points. AI creates the feedback from those notes.AI provides a second opinion
The educator marks the assignment. AI also reads the assignment and rubric and provides its own assessment.AI marks first
AI reads the assignment and highlights strengths, weaknesses and relevant areas of the rubric. The educator then reads the work and determines the mark.AI determines the mark
AI uses the rubric to determine a mark and provide feedback. The educator reads the assignment and checks whether the mark and feedback are appropriate.AI marks and humans spot-check
AI marks the majority of assignments. Educators review borderline cases and a sample of the remaining work.AI does everything
AI reads the assignment, determines the mark and provides the feedback.
Where would you set your limit?
And where do you think your university would set its limit?
Part 2: Is that enough to create a student passport?
In this week's From the Evidence article, we looked at whether AI could use student work collected over a degree to build a kind of student passport — tracking changes in things like argumentation, use of evidence, creativity and communication.
Based on where you drew the line above, would AI still have enough access to student work to create something like that?
If you feel that AI should not be able to read and analyse student work, then probably not. However, if you are comfortable with AI reading and analysing student work — even if it doesn't determine the mark — then perhaps it could.
That raises a broader question:
How much access to student work are we comfortable giving AI, and what could we do with that access beyond marking?