Exploring value judgements in grading: will teachers mark down student work assisted by GenAI, and should they?

Exploring value judgements in grading — a qualitative scenario-based interview study of 33 university teachers in the Greater Bay Area of China (higher education, junior/teaching-track roles across humanities, social, applied, and natural sciences) investigating how they grade student work that may have been assisted by GenAI. The authors find that teachers make value judgements that extend beyond the assignment itself to encompass conjecture about who the student is (person-oriented values: honesty, diligence), what they are capable of (capability-oriented values: independence, GenAI skill, disciplinary mastery), how they relate to others (relation-oriented values: trust), and whether the grading decision leads to good outcomes (justice-oriented values: fairness, beneficence). They foreground validity as a pathway forward and call for greater transparency about how students' GenAI use will be factored into grading decisions, moving beyond the binary "is it cheating?" debate toward the subjectivities of grading in the age of GenAI.

Key Findings

• Grading is inherently a value judgement. Teachers' grading decisions reflect subjective prioritisation of values — often implicit and rarely discussed — rather than purely objective, criteria-based application. Even when rubrics formalise criteria, teachers bring their own interpretations and priorities to them.
• Four value clusters shape grading of GenAI-assisted work. (1) Person-oriented (22 instances): perceptions of student honesty and diligence — GenAI use read as either laziness/shortcut-taking or as diligence/effort depending on the teacher. (2) Capability-oriented (33 instances): what counts as "capability" becomes murky — independence from AI, skilful GenAI use, and disciplinary-specific mastery are weighed differently, with a hierarchy implicitly favouring "completely human" work. (3) Relation-oriented (7 instances): trust built with students over time, and concern that penalising honest disclosure damages teacher-student trust. (4) Justice-oriented (4 instances): fairness (avoiding resentment among students who worked independently) and beneficence (avoiding harm from false accusations).
• Considerable diversity, even conflict, in value judgements. Teachers varied widely; the same teacher could hold competing values (e.g., valuing AI competence while instinctively favouring independent work). Variation is explained at individual (epistemological stages: absolutist/multiplist/evaluativist), disciplinary (humanities more critical of GenAI use; hard/applied sciences fewer grading challenges), and sociocultural (Confucian context valuing hard work and honesty; "AI shaming" stigma) levels.
• Validity as the organising concept. Marking down GenAI-assisted work is justified if and only if the AI use prevented students from demonstrating the outcomes being assessed. Much of the observed value-driven grading introduces construct-irrelevant variance — grades varying on factors (honesty, diligence, trust) irrelevant to the outcomes assessed.
• Transparency is the mitigation. Students and assessors are unclear where the line is between acceptable and unacceptable GenAI use. The paper calls for "two-way transparency" — not just students declaring use, but teachers clarifying how use will impact grades.

Implications for AI in Education

• For Assessment Validity: the study is a direct empirical demonstration of construct-irrelevant variance in human grading of GenAI-assisted work — person-, capability-, relation-, and justice-oriented values all have the potential to introduce variance unrelated to the assessed outcomes. It grounds the validity framing in real teacher practice.
• For Assessment and Academic Integrity: grading is not a binary "cheating vs. not" decision; teachers weigh honesty, diligence, trust, and fairness alongside work quality. This supports the knowledge base's shift from detection toward transparent, validity-preserving assessment design — and shows why disclosure policies must be paired with clarity about grading consequences.
• For Teacher Role and Educational Development: teachers often lack awareness of the value orientations underlying their grading. Professional development should help teachers critically examine their value judgements, align grading with educational goals, and review rubrics to account for GenAI use — targeting the evaluativist stage of epistemological sophistication.
• For Equity In AI Education: value-driven grading risks inequity — e.g., penalising students who honestly disclose GenAI use, or grading down on suspicion of dishonesty. Transparency and consistency in grading are equity concerns, not just validity ones.

Connected Concepts

• Assessment Validity
• Assessment
• Academic Integrity
• Teacher Role
• Generative AI
• Higher Ed
• Trust
• Qualitative Research

Connected Articles

• Farazouli Navigating Uncertainty Teachers GenAI 2026 — university teachers' experiences and perceptions of GAI in teaching and learning (Farazouli et al. 2026)
• Roe AI Humanizers Legitimacy Assessment 2026 — AI humanizers and the performance of legitimacy in assessment (Roe et al. 2026)
• Walton Bearman Assessment Judgement 2025 — judgement in students' work with GenAI on assessment tasks (Walton et al. 2025)
• Chirikov AI Grade Inflation 2026 — AI task displacement as a mechanism of grade inflation (Chirikov 2026)
• Biology Grade Vulnerability GenAI 2026 — vulnerability of biology course grades to AI-mediated dishonesty (Chan et al. 2026)

Citation

Luo, J. (Jess), & Dawson, P. (2026). Exploring value judgements in grading: will teachers mark down student work assisted by GenAI, and should they?. Studies in Higher Education, 51(9), 1970–1984.