AI Marking in Universities: Efficiency Gain or Existential Threat?
As universities increasingly grapple with workloads, some institutions across Australia have permitted staff to use generative artificial intelligence for marking assignments and delivering student feedback. While proponents argue the technology serves as a much-needed second pair of eyes for overburdened educators, critics warn that substituting human judgment with algorithms risks an existential collapse of higher education standards.
Generative artificial intelligence has moved beyond theoretical classroom discussions and entered the grading systems of several Australian institutions. Western Sydney University, the University of Newcastle, Deakin University, RMIT in Victoria, and the University of Adelaide have all introduced qualified allowances permitting academic staff to use AI tools for assessments and administrative feedback workflows.
The policy changes arrive at a critical juncture for higher education. University staff across the sector frequently find themselves pushed to breaking point due to heavy workloads, with grading consuming a substantial portion of their working hours. Proponents suggest that integrating artificial intelligence can help alleviate this administrative strain.
Yet, the adoption is far from uniform. High-ranking institutions including the University of New South Wales, the University of Melbourne, and the University of Sydney have drawn a firm line against utilizing generative AI in the marking process, despite embracing it elsewhere in campus life.
Existential Risks and the Fear of a Slop-Cycle
Armin Alimardani, a senior lecturer in law and technology at Western Sydney University, has studied and experimented with artificial intelligence for years. Speaking from a series of university workshops in Palermo, Italy, Alimardani describes the push toward automated grading as extremely concerning and warns that it poses a serious threat to higher education.
According to critics, allowing algorithms in datacentres to replace human academics creates a potential slop-cycle. In this scenario, students submit assignments generated by artificial intelligence, which are subsequently evaluated and graded by artificial intelligence tools.
Students are already facing significant financial pressures, Alimardani notes. If undergraduates feel they are not receiving proper feedback, or if they suspect their degrees hold less value for employers, the fundamental utility of attending university comes into question.
Safeguards, Verification Drift, and Institutional Policy
Universities permitting automated assistance maintain that strict human oversight remains mandatory. A spokesperson for Western Sydney University emphasized that the institution takes a human-centred approach to generative AI, insisting that the awarding of marks, grades, and feedback must always stay under the direct responsibility of academic staff.
Similarly, the University of Newcastle grants students the right to opt out of AI-assisted grading, while Deakin University permits staff to use AI strictly to improve efficiency and support administrative activities rather than to assign grades directly.
Despite these safeguards, skepticism remains high regarding scalability and oversight. Alimardani points to the phenomenon of verification drift. When 100 academics are given an AI marking tool and instructed to verify its outputs, a fraction will inevitably stop maintaining strict vigilance after observing initial success. As human oversight lapses, artificial intelligence hallucinations risk slipping through unchecked as verified facts.
Such errors mirror recent courtroom scandals involving unchecked algorithmic outputs, which could severely damage university reputations. As the debate continues, institutions such as Monash University, the University of Wollongong, and James Cook University report that they are still evaluating their positions, with AI currently excluded from their active marking procedures.