Australian Universities Split on AI-Assisted Marking as Staff and Students Question Oversight

An Australian student’s assignment may now pass through an AI tool on its way to feedback, even where an academic remains responsible for the result. A Guardian examination found that five universities allow some AI assistance with assessment or feedback; three others bar generative AI from marking. The split raises a harder question than whether the tool is allowed: how closely will a busy lecturer check its work?

Permission stops short of assigning grades

Western Sydney University, the University of Newcastle, Deakin University, RMIT and the University of Adelaide all allow limited AI use in assessment or feedback. Their rules are not identical. Deakin permits AI to support assessment and administrative work, but says staff cannot use it to assign grades. Newcastle gives students the right to opt out of AI use in marking.

Western Sydney says academic staff may use AI for certain parts of assessment and feedback, but must remain responsible for academic judgment, marks, grades and feedback. At the University of NSW, the University of Melbourne and the University of Sydney, the line is stricter: generative AI is barred from marking, though it is allowed in other classroom activities.

responsibility for academic judgment must always remain with university teaching staff

Western Sydney University spokesperson, speaking to the Guardian

The time saved depends on the checking

Armin Alimardani, a Western Sydney senior lecturer in law and technology, sees a possible upside. AI could act as a second pair of eyes, and he is working on models to spot inconsistencies in human assessment. Marking takes up much of academics’ time, he told the Guardian, while many staff are stretched to “breaking point”. He cautioned that the promised benefits of AI assistance remain largely untested.

His concern is “verification drift”: after several AI-generated assessments look sound, an overworked academic might check the next one less carefully and miss a made-up claim. Alimardani offered this as a risk of using the tools at scale, not as a measured failure rate. He worries that errors and weak feedback could leave students doubting whether they were assessed fairly.

That leaves universities with a practical tradeoff. The closer academics review AI suggestions, the less time they may save; the lighter the review, the more they depend on the tool’s judgment. Alimardani questions whether AI can speed up assessment at scale while keeping the safeguards universities promise.

A project email stirs job fears

At Queensland University of Technology, an August staff email described an “assessment uplift project” aimed at making assessment authentic, trusted and scalable in an AI-enabled learning environment. It said broad consultation was underway. The email did not set out a decision to replace human marking, but its broad language alarmed some staff.

A senior QUT academic, speaking anonymously to the Guardian, called the message vague and said staff feared it signaled less human marking. The academic noted that sessional teachers often rely on marking for income. Those are concerns about where the project might lead, not announced staffing changes.

Students want feedback they can trust

The anonymous QUT academic also questioned the message sent to students if universities warn them about integrity problems in using AI for assignments, then use it to assess their work. Alex Cameron, a QUT creative and professional writing student, told the Guardian he wants people to read his writing and give feedback. He said he would not pay $1,000 for a degree in which AI marked his work. His view is a student’s response, not a measure of wider student opinion.

Other institutions have yet to settle on an approach. Monash says AI is not currently used in its marking, while it develops a broader approach to AI in teaching and assessment. The choice facing those still deciding is not simply whether a machine can produce comments. It is what students should be able to expect from the person ultimately responsible for them.