Academic Integrity in the Age of Online Learning

This article is based on research I originally conducted in April 2022 for Saint Mary's University on academic integrity in online courses. I have substantially revised and updated the original research to account for developments in generative artificial intelligence, digital proctoring, privacy and cybersecurity since its completion.

Academic dishonesty is hardly a new problem for universities. What has changed is the environment in which it takes place.

Online learning has given students access to an extraordinary range of educational resources. Recorded lectures, digital libraries, online tutoring and, more recently, generative artificial intelligence have made information and assistance available almost instantaneously. These technologies have considerable educational value. They have also made the boundaries between legitimate assistance, collaboration and academic misconduct increasingly difficult to navigate.

Long before ChatGPT entered university classrooms, students could search repositories for examination questions, upload assignments to answer-sharing platforms or pay someone to complete an assignment on their behalf. Universities responded with plagiarism-detection software, lockdown browsers and remote proctoring.

But research suggests that technology addresses only one part of the problem. Students cheat for complicated reasons. Opportunity matters, but so do academic pressure, perceptions of fairness, institutional culture and whether students believe their peers are cheating. Academic integrity is therefore as much a question of human behaviour and assessment design as technological enforcement.

And as universities adopt more technology to prevent misconduct, another problem emerges: at what point does protecting academic integrity begin to conflict with student privacy, accessibility and cybersecurity?

Academic Misconduct in Canadian Universities

One of the most significant studies of academic misconduct in Canadian higher education was conducted by Julia Christensen Hughes and Donald McCabe. 45% of undergraduate respondents and 29% of graduate respondents reported working with others when an instructor had specifically requested individual work. 38% of undergraduates and 16% of graduate students reported obtaining examination questions or answers from someone who had already taken a test (Christensen Hughes & McCabe, 2006).

The age of the study is important - these figures should not be treated as estimates of how many Canadian students cheat today. What remains relevant is the researchers' observation that some students engaging in unauthorized collaboration, fabrication or falsification did not necessarily consider their behaviour particularly serious.

Digital learning has made these boundaries more complicated. Is discussing an assignment with a classmate collaboration or collusion? Can AI generate an outline but not the final paper? Is uploading an instructor's question to an online tutoring service different from asking a tutor privately?

There may be clear answers within a particular course, but those answers are not necessarily the same everywhere. Universities cannot assume that students intuitively understand where legitimate assistance ends and misconduct begins.

Why Students Cheat

Academic dishonesty cannot be explained simply by dividing students into those who are honest and those who are not.

The fraud triangle, originally developed to understand financial fraud, provides a useful framework. Applied to academic misconduct, it involves three conditions: opportunity, where a student believes cheating is possible without being caught; pressure or incentive, such as academic failure or overwhelming deadlines; and rationalization, where the student finds a way to make the behaviour seem acceptable.

Research has identified similar factors, including feelings of inadequacy, time pressure, attitudes toward cheating and the desire for higher grades (Jordan, 2001; Crown & Spiller, 1998; DeVoss & Rosati, 2002).

Students' perceptions of their peers also matter. If cheating appears widespread, refusing to participate can begin to feel like accepting a competitive disadvantage (Crittenden et al., 2009).

Online learning can intensify some of these conditions. Students may feel less connected to instructors and classmates, particularly in large asynchronous courses, while struggling with material they might have preferred to learn in person. When students feel underprepared while facing high academic stakes, the temptation to look elsewhere for help increases.

From Contract Cheating to Generative AI

By the pandemic, academic misconduct had already become commercialized. Students could search enormous collections of educational material, submit questions to online services or pay another person to complete academic work.

The latter practice, known as ‘contract cheating’, creates a particular problem for plagiarism detection. A custom-written paper may contain little copied material and therefore be original in the technical sense while remaining entirely fraudulent in the academic one (Rigby et al., 2015; Ellis et al., 2018; Lancaster, 2019).

There is also a cybersecurity dimension. Students using questionable academic services may disclose names, payment information, course details or even information relating to university systems. Once personal information or credentials are disclosed to an unknown third party, students have little control over what happens to them.

My original research on this topic was completed in April 2022 . Seven months later, OpenAI publicly released ChatGPT.

Generative AI changed the problem considerably. Students no longer necessarily need an existing answer or another person. AI can generate essays, code, explanations, outlines and solutions within seconds.

Yet describing AI simply as a cheating tool misses the more difficult question it creates.

A student might legitimately use AI to explain a difficult concept, generate practice questions or critique an argument. The same technology could also be used to generate an entire assignment. The important question is therefore not simply whether AI was used, but how it was used, what the assessment was intended to measure, whether its use was authorized and whether it was appropriately disclosed.

Research since ChatGPT's release increasingly points toward clearer institutional guidance, AI literacy and assessment redesign rather than relying solely on detection and prohibition (Xia et al., 2024).

There is also an uncomfortable possibility universities have to confront: if a generic assignment can be completed extremely well by asking an AI system a generic question, perhaps the problem is not exclusively the technology.

It may also be the assessment.

Better Assessment may be the Strongest Defence

Despite increasingly sophisticated cheating technologies, one of the most consistently recommended interventions remains surprisingly simple: talk to students about academic integrity.

Policies buried in course outlines are insufficient if students do not understand how to apply them. Students should know whether collaboration is permitted, whether AI can be used for brainstorming or editing, whether online resources are allowed during an assessment and whether course material can be uploaded to external services.

Assessment design matters just as much.

Generic questions invite generic answers. Assignments grounded in a particular class discussion, dataset, case study or real-world problem require greater engagement with the course itself. Quantitative questions can ask students not only to calculate an answer but explain its meaning, assumptions or limitations.

There is also considerable value in scaffolding. Instead of seeing only a final paper, instructors can review the research question, preliminary sources, argument, outline or draft at different stages. This makes complete outsourcing more difficult while giving struggling students opportunities to seek help before reaching a point of desperation.

High-stakes assessments deserve similar scrutiny. Very short response windows, preventing backtracking or imposing restrictive technological controls may make certain forms of cheating harder, but they can also disadvantage honest students, particularly those requiring accommodations.

Making an assessment harder to cheat on is useful only if it remains a fair and valid measure of what students are supposed to know.

The Problem with Digital Proctoring

Remote examinations created enormous demand for lockdown browsers and automated or live proctoring. These technologies can make some forms of cheating more difficult, but they cannot make a student's home equivalent to a controlled examination hall.

A lockdown browser controls one device. A webcam sees only part of a room. Automated systems can identify unusual behaviour without knowing why it occurred. Looking away from a screen is an observation - it is not proof of cheating.

This distinction matters because proctoring technology can involve highly sensitive personal information, including students' faces, voices, identification, screens, IP addresses and images of their homes. Some systems may also process biometric information or use automated analysis to evaluate behaviour.

Universities have a legitimate interest in protecting their assessments. Students do not surrender their privacy rights simply because an examination has moved online.

In 2024, Ontario's Information and Privacy Commissioner investigated McMaster University's use of Respondus exam-proctoring software. While recognizing the legitimate academic-integrity purpose of online proctoring, the Commissioner identified concerns involving notice to students, use of personal information and contractual safeguards.

The case demonstrates why proctoring software is not merely an educational tool. It is also a data-governance and cybersecurity system.

Universities evaluating these technologies should therefore ask what information is collected, why it is necessary, where it is stored, who can access it, how long it is retained, whether biometric information is involved, what third parties receive it and what happens following a security breach.

Accessibility must be considered at the same time. Not every student has a modern computer, high-speed internet, a private room or the ability to remain continuously visible to a webcam. Others rely on assistive technologies or accommodations that can conflict with rigid proctoring requirements.

Security controls should not inadvertently turn ordinary student behaviour into suspicious behaviour.

What Universities are Really Trying to Assess

There is an understandable appeal to technological solutions for academic misconduct. A university can purchase a lockdown browser, install plagiarism software, require webcams and generate reports. Changing assessment design and academic culture is considerably harder.

But no single technological control can solve academic dishonesty.

The stronger approach resembles a familiar principle in cybersecurity: defence in depth. Universities can combine clear expectations, academic-integrity education, meaningful assessment design, reduced dependence on unnecessarily high-stakes examinations and appropriate technological controls. Where monitoring is necessary, it should be accompanied by human review, privacy safeguards, accessibility considerations and procedural fairness.

Generative AI makes this approach even more important because it forces universities to ask what they are actually assessing.

There are legitimate reasons to require students to write independently, calculate without automated assistance and develop foundational skills before relying on technology. There will also be circumstances where graduates are expected to use AI professionally. In those cases, knowing when AI is useful, recognizing when it is wrong, protecting confidential information, identifying bias and taking responsibility for the final result may themselves become skills worth assessing.

Academic integrity should not become a contest between increasingly sophisticated ways of obtaining unauthorized assistance and increasingly sophisticated ways of detecting it. That contest has no obvious endpoint.

The more durable question is whether an assessment gives students a reason to engage intellectually with the material and gives instructors credible evidence that learning has taken place.

Technology can help protect that process. It cannot replace it.

References

Christensen Hughes, J. M., & McCabe, D. L. (2006). Academic misconduct within higher education in Canada. Canadian Journal of Higher Education, 36(2), 1–21.

Crittenden, V. L., Hanna, R. C., & Peterson, R. A. (2009). The cheating culture: A global societal phenomenon. Business Horizons, 52, 337–346.

Crown, D. F., & Spiller, M. S. (1998). Learning from the literature on collegiate cheating: A review of empirical research. Journal of Business Ethics, 17, 683–700.

DeVoss, D., & Rosati, A. C. (2002). "It wasn't me, was it?" Plagiarism and the web. Computers and Composition, 19, 191–203.

Ellis, C., Zucker, I. M., & Randall, D. (2018). The infernal business of contract cheating: Understanding the business processes and models of academic custom writing sites. International Journal for Educational Integrity, 14.

Jordan, A. E. (2001). College student cheating: The role of motivation, perceived norms, attitudes, and knowledge of institutional policy. Ethics & Behavior, 11(3), 233–247.

Lancaster, T. (2019). Social media enabled contract cheating. Canadian Perspectives on Academic Integrity, 2(2), 7–24.

Rigby, D., Burton, M., Balcombe, K., Bateman, I., & Mulatu, A. (2015). Contract cheating and the market in essays. Journal of Economic Behavior & Organization, 111, 23–37.

Xia, Q., Weng, X., Ouyang, F., Lin, T. J., & Chiu, T. K. F. (2024). A scoping review on how generative artificial intelligence transforms assessment in higher education. International Journal of Educational Technology in Higher Education, 21, 40.