💻 computer science

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

This study reveals that despite identical institutional policies, Canadian university students perceive generative AI use in computing education as significantly more unethical than their South Korean counterparts, a disparity attributed to cultural differences in individualism, power distance, and uncertainty avoidance that underscores the need for culturally responsive AI guidelines.

Original authors: Brian Harrington, Irina Zlotnikova, Gayathri Nadarajan, Samuel Ekundayo

Published 2026-07-23

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Brian Harrington, Irina Zlotnikova, Gayathri Nadarajan, Samuel Ekundayo

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are walking into a giant, global classroom where the teacher has just handed out a new, magical tool: a robot that can write code, solve math problems, and draft essays in seconds. This tool is called Generative AI (GenAI). It's like having a super-smart ghostwriter who can finish your homework instantly. But here's the big question: Is using this ghostwriter cheating, or is it just smart studying?

For a long time, schools have had strict rules about "academic integrity." Think of this as the school's honor code. It says, "What you submit must be 100% yours." If you copy someone else's work, that's plagiarism, and it's a big no-no. But this new robot changes the game. It blurs the line between "I did this" and "The robot did this." Now, educators and students everywhere are arguing: Is it okay to ask the robot for help? Does it matter if you understand the code it wrote? And does the answer change depending on where you live? This is the puzzle that researchers are trying to solve. They want to know if a student in Canada sees this robot differently than a student in South Korea, and why.

The Great "Alice" Experiment

To figure this out, a team of researchers from Canada, Botswana, New Zealand, and South Korea decided to play a game of "What Would You Do?" They created a story about a student named Alice.

They didn't just ask students, "Is AI cheating?" Instead, they gave them eight different scenarios, like little movie clips, showing Alice using the AI tool in various ways. In some stories, Alice used the AI to just get a tiny idea for a project. In others, she let the AI write almost the entire code for her. In some, she understood exactly what the AI did; in others, she just copy-pasted the result without knowing how it worked.

Then, they asked students from two very different places to judge Alice:

Canada: Representing a culture that values individual achievement and personal originality.
South Korea: Representing a culture that often values group harmony, respect for authority, and shared knowledge.

The researchers gave the students a simple question for each story: "Did Alice act unethically?" and "Was what Alice did against the rules?"

The Big Surprise: It's All About Culture

Here is the twist: Even though the schools in Canada and South Korea had exactly the same rules about AI (or at least, no clear rules either way), the students judged Alice very differently.

The Canadian students were much stricter. They were more likely to say, "Yes, Alice cheated!" and "Yes, that's against the rules!" The South Korean students were more lenient. They were more likely to say, "Well, maybe it's okay," or "It depends."

It's like two groups of people looking at a painting. One group says, "That's a fake! The artist didn't paint it!" The other group says, "It doesn't matter who painted it; it's still a beautiful picture." The paper suggests that this isn't because the Korean students are "bad" at following rules, but because their culture views knowledge differently. In many collectivist cultures (like South Korea), sharing resources and helping each other succeed is seen as a virtue. In individualist cultures (like Canada), the focus is on your unique contribution and your personal effort.

What Made the Biggest Difference?

The researchers also looked at what Alice actually did in the stories to see what made the students angry. They found one thing that made everyone, everywhere, more likely to call it cheating: The Amount.

When Alice used the AI to write a huge chunk of the code (like the whole main function), both Canadian and Korean students were much more likely to say, "Whoa, that's too much!" However, even when the amount of AI code was the same, the Canadian students still thought it was more unethical than the Korean students did.

The study also checked if other things mattered, like:

When she used the AI (at the start, middle, or end of the project).
How well she understood the code she got.

While these factors did change the answers a little bit, the amount of AI-generated code was the biggest driver of whether students thought it was wrong.

Why Does This Matter?

The paper suggests that we can't just make one set of rules for the whole world and expect everyone to agree. If a university in Canada says, "AI is cheating," and a university in Korea says, "AI is a helpful study buddy," students might get confused.

The researchers used a famous theory called Hofstede's Cultural Dimensions to explain this. They compared the two countries using concepts like:

Individualism vs. Collectivism: Do we focus on "Me" or "Us"?
Power Distance: Do we question the teacher's rules, or do we follow them blindly?
Uncertainty Avoidance: Do we like clear rules, or are we okay with a bit of gray area?

They found that these cultural "personality traits" shape how students think about right and wrong. Canadian students, coming from a low-power-distance, individualist culture, felt a strong personal responsibility to be original. Korean students, coming from a higher-power-distance, collectivist culture, were more focused on the group's success and the authority of the institution. If the school didn't explicitly say "No AI," they were more likely to assume it was okay to use.

The Bottom Line

This study didn't prove that one culture is right and the other is wrong. Instead, it showed that culture is the lens through which we see ethics.

The researchers are careful to say that their findings are based on what students said they would do in a survey, not necessarily what they actually did in real life. But the results are clear: We need to stop assuming that everyone sees "cheating" the same way. To make AI work fairly in schools around the world, we need rules that respect these cultural differences while still keeping the core idea of honesty alive. It's not just about banning or allowing robots; it's about understanding the human hearts and minds behind the screens.

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

Problem Statement

The rapid integration of Generative AI (GenAI) into higher education has precipitated urgent debates regarding academic integrity, particularly within computing education where code generation is a core task. While institutions are grappling with how to regulate these tools, a critical gap exists in understanding how cultural contexts shape student perceptions of what constitutes ethical versus unethical AI use. Existing literature often treats academic integrity norms as universal or focuses on single-country contexts, failing to account for how deep-seated cultural values (such as individualism vs. collectivism) influence the interpretation of "plagiarism" in the age of AI. This study addresses the lack of cross-cultural empirical data comparing student judgments of GenAI use in computing coursework, specifically examining whether students from different cultural backgrounds perceive identical AI-assisted practices differently, even when institutional policies are functionally identical.

Methodology

The researchers employed a quantitative, scenario-based survey design administered in Fall 2024 to students at two institutions: the University of Toronto Scarborough (CAN) and Sungkyunkwan University (KOR).

Participants: The study surveyed 262 students from CAN (62% response rate) and 48 students from KOR (74% response rate). The CAN sample consisted of non-major introductory programming students, while the KOR sample included students from data science-related courses (relational databases, NLP, and ethical/social impact of data).
Experimental Design: The survey utilized eight distinct vignettes (scenarios) describing a student named "Alice" using GenAI tools in a programming assignment. Each scenario was systematically varied across three orthogonal factors:
1. Amount: The volume of AI-generated code in the final submission (Low, Medium, High).
2. Understanding: The level of student comprehension demonstrated regarding the generated code (Low, Medium, High).
3. Stage: The point in the development lifecycle where the tool was used (Early, Middle, Late).
Measurement: For each scenario, participants responded to two prompts on a five-point Likert scale:
1. "Alice acted unethically in this scenario."
2. "What Alice did would be against the rules in my institution."
Contextual Controls: Both institutions had identical, instructor-discretion-based policies regarding GenAI at the time of the study, with no explicit institution-wide mandates. This controlled for policy variance, isolating cultural perception as the primary variable.
Statistical Analysis: The study utilized Mann-Whitney U tests to compare ordinal responses between the two countries. A two-way ANOVA was used to assess the impact of country and year of study, and a repeated-measures ANOVA analyzed the influence of the scenario factors (Amount, Understanding, Stage).

Key Results

The analysis revealed statistically significant cross-cultural differences in ethical judgments, despite the functional identity of institutional policies.

Cultural Divergence: Canadian students were consistently more likely to perceive GenAI use as both unethical and against institutional rules compared to Korean students across nearly all scenarios. The mean rating for "Acted unethically" was 2.6 for CAN students versus 1.78 for KOR students (on a 5-point scale). Similarly, the mean for "Against the rules" was 2.81 (CAN) versus 1.80 (KOR).
Dominant Factor: The "Amount of AI-generated code" was the most significant predictor of ethical judgment. Scenarios involving high or medium amounts of copied code elicited the strongest divergence between the two groups. While the "Understanding" and "Stage" factors were statistically significant, their effect sizes were smaller than that of the code volume.
Correlation: There was a strong positive correlation ( $r = 0.85$ ) between a student's perception of unethical behavior and their perception of rule violation, suggesting that students often project personal ethical standards onto institutional expectations.
Non-Significant Variables: Academic progress (year of study) and general opinions on the negative impact of GenAI on academic integrity or learning quality showed no statistically significant differences between the two populations.

Key Contributions

Empirical Evidence of Cultural Bias in AI Ethics: The study provides robust statistical evidence that cultural background significantly influences the ethical evaluation of GenAI use in computing education. It demonstrates that "identical" policies are interpreted through distinct cultural lenses, with individualistic cultures (Canada) exhibiting stricter adherence to originality norms than collectivist cultures (South Korea).
Factor-Specific Sensitivity: The research identifies that the volume of AI-generated content is the primary driver of ethical concern, more so than the stage of use or the student's demonstrated understanding. This suggests that policy frameworks focusing solely on "how" AI is used may miss the critical variable of "how much" is used.
Theoretical Application: The findings are interpreted through Hofstede's cultural dimensions framework. The results align with theoretical expectations:
- Individualism vs. Collectivism: Canadian students (individualistic) viewed AI use as a violation of personal intellectual property, whereas Korean students (collectivist) were more likely to view it as a shared resource or collaborative aid.
- Power Distance: The higher power distance in Korea may lead students to defer to authority or assume permissiveness unless explicitly prohibited, whereas low power distance in Canada encourages questioning and strict adherence to self-imposed ethical standards.

Significance and Claims

The paper claims that equitable integration of GenAI in education cannot rely on a "one-size-fits-all" policy approach. The authors argue that institutional guidelines must be culturally responsive. They posit that while fundamental principles of academic honesty must be upheld, the interpretation of what constitutes a violation is mediated by cultural values such as individualism, collectivism, and uncertainty avoidance.

The study concludes that future policy development and pedagogical interventions must acknowledge these cultural nuances. Specifically, the authors advocate for the creation of nuanced AI-use guidelines that are sensitive to local cultural contexts while maintaining core academic integrity standards. They emphasize that without such cultural responsiveness, institutions risk misinterpreting student behavior or failing to effectively educate diverse student populations on the ethical use of emerging technologies. The authors modestly note that their findings are limited to two specific institutions and call for broader, longitudinal, and mixed-methods research to further validate these cross-cultural patterns.