Did Alice Do Wrong? Cross-Cultural Differences in Student Perceptions of Generative AI Use in University Computing Education

Synthesis: A scenario-based survey (Fall 2024) comparing how computing students at Canadian and South Korean universities judged the ethicality and policy compliance of AI-assisted coding practices. Despite functionally identical institutional policies, Canadian students were consistently and significantly more likely to rate GenAI use as unethical and against the rules (Mann-Whitney U tests across nearly all scenarios). Culture, not policy text, drove the perceived wrongness of identical behaviors.

The result complicates the knowledge base's Academic Integrity thread: policy harmonization does not produce perception harmonization, so multi-national or culturally diverse cohorts will interpret the same rules differently — an Equity concern for enforcement and grading. It extends Students' Perception Accuracy of Partners' AI Use and its Relation to Collaboration Performance and The impact of generative artificial intelligence on academic development of Chinese students in humanities and social sciences with direct cross-national comparison in computing education, and argues that A Comparative Analysis of Institutional and Course Generative AI Policies within Higher Education: Implications for Instruction in Computing Education need culturally aware communication, worked examples, and scenario-based clarification rather than abstract rule statements.

What this means for practice

• Instructors. Teach the rules with scenarios rather than abstractions: despite functionally identical institutional policies, Canadian students rated the same AI-assisted coding behaviors as unethical and against the rules significantly more often than South Korean students across nearly all eight scenarios.
• Instructors. Spend class time on the factor that drove judgments most — the share of AI-generated code incorporated into an assignment — and show worked examples that mark where assistance stops being acceptable in CS Education tasks.
• Faculty developers. Build cultural literacy into teaching-assistant and instructor onboarding so that grading and Academic Integrity conversations do not assume one cohort's reading of the policy is the only legitimate one.
• Administrators. Treat policy harmonization as insufficient and check enforcement for cultural bias, since identical behavior judged differently across a diverse cohort is an Equity problem rather than a compliance one.
• Administrators. Re-issue GenAI guidance with concrete scenarios and review it regularly, pairing it with training that develops critical judgment about AI use instead of publishing prohibition lists.

Limitations

• Two universities only, one in Canada and one in South Korea, so variation within each country across regions, disciplines and institution types is not captured.
• Sample sizes were disproportionate: 262 students at the Canadian institution (62% of a 420-student course) against 48 at the Korean institution (74% of 65); expanding the Korean sample was not feasible without breaking comparability.
• Judgments were self-reported responses to hypothetical scenarios, so they may differ from real coursework behavior and are subject to social desirability bias and differing item interpretation across cultures.
• Scenario-level contrasts may partly reflect cross-cultural response-style differences such as neutral-category use, so the Mann-Whitney U results should not be read as pure median shifts.

Citation

Harrington, B., Zlotnikova, I., Nadarajan, G., & Ekundayo, S. (2026). Did Alice Do Wrong? Cross-Cultural Differences in Student Perceptions of Generative AI Use in University Computing Education. arXiv preprint (cs.CY).