AI grading means university assessments could theoretically never be read by a human

The use of artificial intelligence (AI) for grading university assessments raises two possibilities: a dystopian future where AI handles all grading without human intervention, or an idealistic scenario where AI frees educators to focus on teaching and research. In reality, AI grading has been commercially available since 1998, with Pearson's systems scoring over 39 million responses.

AI models, like ChatGPT, predict grades by comparing student work to vast databases of texts. Universities generally avoid openly discussing their AI grading methods, but they must comply with strict privacy regulations. Some institutions, like Deakin University, prohibit teachers from uploading student data to public AI platforms, fearing exposure of private data.

While AI grading can standardize feedback and reduce grading burdens, current systems are not reliable enough for sole reliance. A study found that ChatGPT overestimated student essays compared to human graders and struggled with specific grading criteria. Additionally, AI grading could distance students from teachers, as students may turn to AI for feedback first.

Research suggests that removing numerical grades might benefit student-teacher relationships and encourage independent thinking. Universities like the University of New South Wales have experimented with removing numerical scores and evaluating students based on competency against defined outcomes. Ultimately, the future of grading may not be an either-or situation but rather a reevaluation of assessment systems to prioritize meaningful feedback and human interaction.