Who Should Grade My Work? Student Perspectives on Transparent AI-Assisted Writing Assessment in Higher Education

The study "Who Should Grade My Work? Student Perspectives on Transparent AI-Assisted Writing Assessment in Higher Education" (AlGhamdi, 2026) investigates student reactions to transparent AI-assisted writing assessment, where learners are explicitly informed that ChatGPT generated their feedback and scores.

Key findings from the qualitative inquiry involving undergraduate computing students include:

• Distinction Between Utility and Authority: Students treat feedback utility and evaluative authority as separate judgments. They accept GenAI feedback as useful for surface-level revision but consistently position the human instructor as the appropriate authority for final grading decisions.
• Awareness of Limitations: Participants demonstrated awareness of AI’s contextual and pedagogical limitations, leading to conditional trust in the system’s ability to evaluate complex academic work.
• Role of Human Instructors: The study highlights a reflection on the institutional role of human instructors, with students viewing human evaluation as essential for maintaining the integrity of the assessment process, despite the practical benefits of AI-generated comments.

The research suggests that while students value the immediate and specific feedback provided by GenAI, they do not view it as a legitimate substitute for human judgment in determining final academic grades.

This material examines student perspectives on transparent AI-assisted writing assessment in higher education, focusing on cases where learners are explicitly told that writing feedback was generated by a generative AI system. Rather than treating AI as a hidden backend or a neutral grading tool, the study foregrounds the social and evaluative question posed by the title: who should grade my work? It investigates how students interpret the authority, reliability, and pedagogical value of GenAI feedback when its non-human provenance is made visible, and how that transparency shapes their willingness to accept, question, or act on the feedback.

A key insight is that student responses are not simply acceptance or rejection of AI-generated assessment. Students appear to evaluate AI feedback along dimensions such as specificity, tone, consistency, perceived expertise, and alignment with learning goals. Transparency can reduce deception or overtrust, but it may also lead students to discount feedback that would otherwise be useful, especially when they view grading as inherently human, judgmental, or socially accountable. The work therefore contributes a more nuanced account of AI-assisted assessment: the issue is not only whether AI can produce plausible feedback, but how learners make sense of AI as an assessor and what that means for feedback uptake, revision behavior, and perceived fairness.

The material matters because it speaks directly to responsible deployment of LLMs in educational assessment. It suggests that transparency is necessary but not sufficient; effective AI-assisted grading also requires clear role boundaries, high-quality feedback design, opportunities for human review or appeal, and explicit communication about what AI feedback can and cannot determine. For institutions, instructors, and researchers, the study highlights the need to treat GenAI feedback as a pedagogical and ethical intervention, not merely an efficiency tool, especially in high-stakes contexts where grading carries consequences for students’ grades, confidence, and academic progression.