Cheating or not cheating? Rethinking AI-giarism and academic integrity through secondary students' ethical reasoning

Synthesis: Chan (2026) qualitatively analyzed open-ended responses from 133 Hong Kong secondary students (aged 14–16) to show that reasoning about the ethicality of generative AI in homework resists simple "cheating or not cheating" binaries. Applying Waltzer and Dahl's moral-developmental framework, the study found students engage in conditional, context-sensitive ethical reasoning about purpose of use, learning impact, intent, and fairness, positioning themselves as active moral agents rather than passive rule-followers. The findings challenge fixed notions of AI-giarism and argue that Academic Integrity, Assessment design, and AI Literacy education must foreground ethical reasoning, transparency, and student agency.

Key Findings

1. Conditional use dominates. Most secondary students evaluated AI use in a maths homework scenario as acceptable only under conditions—acceptable for inspiration, clarification, reference, or checking, but not for producing final answers or direct copying.
2. Distinct ethical sub-criteria. Students weighed three factors in deciding whether AI use was cheating: purpose of use (62.4%), impact on learning (54.1%), and student intent (51.9%).
3. Ethical reasoning beyond rules. Students judged acceptability through academic norms/conventions (48.1%), comparison with other tools like Google (31.6%), and accountability/consequences including fairness (23.3%).
4. Intent as the moral boundary. Students consistently framed intent—learning-oriented Help-Seeking versus effort-avoiding substitution—as the decisive factor distinguishing legitimate support from dishonesty.
5. AI as a moral agent, not a fixed category. Roughly 9% of the sample rejected all AI use as undermining independent thinking; the majority treated academic misconduct as a negotiated moral space shaped by intent and condition rather than a fixed category.
6. Fairness and equity concerns. Students worried AI could "trick the teacher" into overestimating understanding, disadvantage capable peers, or—alternatively—democratize access to help often unavailable without private tutoring.
7. Emerging norms of transparency. Several students proposed disclosing AI use through footnotes or references, suggesting their ethical reasoning may run ahead of existing institutional policy.

What this means for practice

• Instructors. Write school-level AI policy around the conditions students themselves weigh — purpose of use (62.4%), impact on learning (54.1%), and student intent (51.9%) — instead of rigid bans or binary cheating definitions.
• Instructors. Teach ethical reasoning and self-regulation rather than compliance: students already reasoned conditionally, so the work is helping them articulate why a particular use is problematic.
• Instructors. Do not rely on detection-led responses to AI-giarism: ethical status depends on intent and use, not on observable output.
• Instructors. Move assessment toward process-visible, reflective, AI-resilient formats that require justification of choices and evaluative judgment, and involve students in integrity policy development so commitment and Trust are shared rather than imposed.
• Learners. Push for transparency norms: several students proposed disclosing AI use in footnotes or references, and roughly 9% rejected all AI use outright — positions that run ahead of existing institutional policy.

Limitations

• The data are 133 students' open-ended written responses (approximately 8,116 words, averaging about 80 words each) to a hypothetical maths homework scenario, not observed behavior in assessed tasks.
• Participants were recruited by convenience sampling from only seven Hong Kong secondary schools, all enrolled in a non-graded 12-hour AI literacy course, so their responses likely over-represent ethical awareness relative to students without that exposure.
• The Padlet format constrained response length and depth and allowed no follow-up probing, so some responses may reflect initial judgments rather than elaborated ethical positions.
• The findings are tied to one scenario: students' stated reasoning may differ from their behavior in high-stakes assessment, where performance pressure plays a greater role.

Citation

Chan, C. K. Y. (2026). Cheating or not cheating? Rethinking AI-giarism and academic integrity through secondary students' ethical reasoning. Computers & Education, 253, 105698.