The stakes are considerable. Universities in Australia and New Zealand collectively host close to one million international students and are widely regarded as global leaders in the adoption of emerging technologies. How these institutions regulate generative artificial intelligence, or GenAI, carries implications well beyond the region. Yet until now, GenAI policies in the higher education systems of both nations have remained largely unexamined, even as surveys of academic staff reveal growing anxiety about how, or whether, to integrate the technology into teaching.

To map the landscape, the researchers deployed a three-pronged methodological strategy. First, they carried out a rapid bibliometric analysis of scholarly literature, searching the Scopus database on February 1, 2025, for journal articles published in English in 2023 and 2024 that combined terms relating to generative AI, policy, and higher education. The initial search returned 164 records, which were screened down to 91 articles. The software package VOSviewer was then used to generate visual maps of keyword co-occurrence, revealing the intellectual structure of a research field that grew fivefold in a single year, from just 15 publications in 2023 to 76 in 2024, a surge that tracks closely with the public release of ChatGPT in November 2022.

The bibliometric analysis revealed a field dominated by the social sciences, which accounted for 75 of the publications, followed by computer science with 34. Authors came from 41 nations, with the strongest contributions from the United States, the United Kingdom, Hong Kong, Australia, Singapore, and Saudi Arabia. Keyword mapping identified four distinct thematic clusters: higher education, students, artificial intelligence in education, and GenAI itself. The most prominent keywords were “higher education,” with 46 occurrences, “generative AI” with 38, “ChatGPT” with 35, and, tellingly, “academic integrity” with 19. At the other end of the spectrum, terms such as “Bloom’s taxonomy,” “pedagogy,” and “holistic competencies” barely registered, suggesting that scholarly attention has concentrated on integrity risks and technological capability rather than on the deeper pedagogical questions GenAI raises. International students, despite being among the most affected groups, were also conspicuously under-represented in the literature.

The second and third strands of the study turned from the scholarly record to institutional practice. The researchers compiled a list of all fifty public and private universities in Australia and New Zealand and, working independently, searched each institution’s website between August 23, 2023, and February 25, 2024, for policies addressing artificial intelligence or generative AI. One institution, Murdoch University, was excluded because its policies are not publicly accessible, leaving 49 universities in the analysis. Each relevant policy was then subjected to a qualitative content analysis, in which policy text was coded as prohibitive, moderative, or encouraging in its stance toward GenAI. To guard against bias and ensure consistency, the coding scheme was piloted on a subset of documents, refined through iterative memoing, documented in an audit trail, and validated through peer debriefing with a second researcher.

The results of the content analysis were striking in their uniformity and their vagueness. Of the 49 universities, only 22 had policies that explicitly mentioned GenAI, and of those, just one, the University of Technology Sydney, possessed a policy dedicated specifically to artificial intelligence, its “Artificial Intelligence Operations Policy.” Every other institution folded its GenAI references into broader documents on academic integrity, student misconduct, responsible research, or assessment standards. The overwhelming majority adopted what the researchers term a moderative stance: GenAI use was neither banned nor embraced, but made conditionally acceptable, permitted when authorised by an academic, or when its use was appropriately acknowledged. Macquarie University’s academic integrity policy, for example, defines unauthorised use as occurring when a student submits material produced by generative artificial intelligence as their own work. Massey University in New Zealand stipulates that AI tools may not be used to generate summative assessment tasks that are then “uncritically submitted” as the student’s own work, unless the assessment criteria explicitly allow it.

Beneath this moderation, however, lay a conspicuous gap: apart from Griffith University and the University of Western Australia, both of which require a citation, the policies offered almost no guidance on what constitutes due acknowledgement of AI use. Prohibitive language was rarer and often blunt. The University of New England classifies presenting AI-written work under one’s own name as a breach of academic integrity, while Victoria University lists generating examination responses via an AI model as cheating. Some policies were more equivocal, warning that AI use “may” constitute misconduct without specifying when it is or is not permitted, as in the policies of Central Queensland University and Flinders University. Only a handful of institutions, most notably the University of Technology Sydney, took an encouraging tone, describing how AI can support efficiency in operations, provide feedback to students, and help identify at-risk or high-achieving learners. Many universities, moreover, mixed stances within a single document, prohibiting AI-generated submissions in one clause while carving out exceptions with “specific permission” in another.

To triangulate these findings, the researchers ran a lexical analysis of the policy texts using Leximancer, a data-mining program that applies Bayesian reasoning to detect which words and concepts travel together in a corpus. The resulting concept map revealed four dominant themes: “artificial,” “intelligence,” “include,” and “support.” The terms “artificial” and “intelligence” each achieved a relevance percentage of 100 percent, making them the most recurrent concepts in the entire corpus, followed closely by “use” and “assessment,” each at 91 percent. What the map did not contain was perhaps the most significant finding of all: despite being housed within integrity-related policy frameworks, the concept of “integrity” never clustered strongly enough to form a theme. The discourse surrounding GenAI, in other words, is operational and procedural, centred on what students may or may not do with the technology in assessment contexts, rather than grounded in the ethical principles that academic integrity policies are ostensibly designed to uphold. When universities write about artificial intelligence, they write about assessment risk, not about honesty, transparency, or the values underpinning scholarship.

Taken together, the three analyses paint a sector that has prioritised academic integrity and ChatGPT in its scholarship, adopted a cautious, conditionally permissive stance in its policies, and yet failed to articulate the integrity language that would give those policies ethical coherence. The authors argue that this incoherence matters because the prevailing focus on preventing, detecting, and evaluating GenAI use is likely to prove inadequate on its own. Plagiarism detection software, after all, remains unreliable at identifying AI-generated text, and wrongful accusations have already fallen disproportionately on international students, some of whom have been mistakenly flagged for using GenAI, while mature students have been largely overlooked in policy discussions altogether. Since GenAI literacy combined with informed judgement is what enables responsible use, the researchers contend that enforcement-focused approaches should be complemented by efforts to build academic capability and foster a genuine culture of integrity.

The study also carries concrete recommendations for university administrators. The authors call for clearer, more consistent definitions of authorised and unauthorised GenAI use, illustrated with discipline-specific examples to reduce ambiguity across subjects and assessment types. They urge institutions to embed explicit integrity-related terminology within GenAI policies, so that the ethical foundations of responsible use are communicated rather than implied. They recommend structured educational supports, including training modules and exemplars of appropriate tool use, arguing that policies should be adaptive rather than solely restrictive. And they make the case for student-centred policy design: universities should engage students directly as co-designers, because policies developed without student involvement risk unintended consequences for institutional inclusiveness and student wellbeing. A bottom-up, emergent approach, the authors suggest, is likely to foster greater uptake, compliance, and shared understanding than top-down universal mandates, which can also place heavy pedagogical demands on academic staff.

For researchers, the findings open a rich agenda: evaluating whether current policy directions support or inhibit teaching innovation, assessing the disciplinary-level effectiveness of GenAI policies, and examining the professional development that academics have so far been denied. The authors acknowledge limitations, including reliance on a single bibliometric database, the exclusion of non-English publications, and the fact that publicly available policies reveal nothing about how they are implemented in practice. Even so, their conclusion is clear. Responsible GenAI policies across Australia and New Zealand, and by extension elsewhere, must be adaptable, stakeholder-oriented, and pedagogically fit for purpose. As generative AI continues to evolve at a pace that outstrips institutional governance, the universities that thrive will be those that move beyond the binary of prohibition and permission, and instead rebuild their policies, and their assessments, around meaningful engagement with the students the policies are meant to serve.

Dr. MacNutt’s recent paper, titled “Reflective Writing Assignments in the Era of GenAI: Student Behavior and Attitudes Suggest Utility, Not Futility,” was published in the esteemed journal Advances in Physiology Education. The study spans nearly 400 participants enrolled in three different health and exercise science courses, all of whom were surveyed anonymously about their use of AI tools during at least five reflective writing assignments. Each course operated under a unified AI policy allowing students the option to incorporate GenAI in their work, offering a controlled backdrop to better understand how students interact with these emergent technologies.

As GenAI tools such as ChatGPT represent the cutting edge of natural language processing, enabling users to communicate with complex language models that can generate human-like text responses, their impact on educational practices continues to be a topic of intense debate. Dr. MacNutt explains that while these tools hold incredible promise for enriching student learning experiences, they also present potential risks if exploited to bypass genuine intellectual engagement. Hence, the concern among educators and institutions revolves largely around maintaining academic integrity while embracing technological innovation.

The findings from this UBCO study are particularly illuminating in this regard. Approximately one-third of the surveyed students reported actively using GenAI tools, but the motivations driving this usage defy the simplistic narrative of cheating or laziness. Over 80% of those who used AI cited reasons that span from expediting task completion to fostering deeper learning, and striving for high academic achievement. This nuanced perspective suggests that students perceive AI as a means to support their educational goals rather than circumvent them.

More specifically, the data reveals that the predominant use of AI was to initiate the composition process or to assist in revising and polishing sections of their assignments. It is crucial to highlight that only a minuscule 0.3% of assignments were predominantly authored by AI, indicating that most students maintain active authorship and critical engagement with their work. This challenges the often alarmist view that AI tools are being employed to produce entire pieces of academic writing with minimal human involvement.

An intriguing aspect that emerges from this research is the ethical dimension underlying student behavior toward AI tools. Participants conveyed that their usage choices were guided by a personal sense of ethical responsibility and a genuine desire to leverage AI in ways that scaffold learning rather than undermine it. This finding disrupts common assumptions that undergraduate learners prioritize grades above learning, instead painting a picture of a conscientious and discerning student body adapting to new educational technologies.

The study also underscores complexities relating to equity and access. Dr. MacNutt notes that paid versions of GenAI platforms may offer superior functionality, potentially giving an unfair advantage to students who can afford these subscriptions. This raises important questions about digital divides and the creation of new forms of educational inequality that mirror broader societal disparities. Ensuring equitable access to AI resources remains a pressing challenge for institutions aiming to incorporate these tools fairly.

Moreover, GenAI’s utility for students with diverse needs cannot be overstated. Individuals learning English as an additional language or those with reading and writing disabilities stand to benefit significantly from AI-assisted writing support. Such applications highlight the potential for AI to enhance inclusivity and personalized learning within higher education, provided ethical frameworks and support structures accompany technological adoption.

The implications of this study are far-reaching for educators and policy makers tasked with navigating the evolving landscape of technology-enhanced learning. Dr. MacNutt advocates for policies that emphasize collaboration between students and AI rather than surveillance or punitive measures. This approach calls for fostering digital literacy and ethical AI use, enabling learners to harness these tools responsibly and strategically within their academic journeys.

As AI technology continues to advance with rapid innovation cycles, ongoing research will be critical to monitor changing student behaviors, attitudes, and outcomes. The UBCO study offers a foundational understanding but also serves as a call to action for universities worldwide to develop responsive educational frameworks that integrate GenAI effectively while safeguarding academic integrity.

Ultimately, this research reorients the conversation away from alarmism and toward a balanced recognition of both the opportunities and challenges that generative AI brings to education. By appreciating students’ nuanced engagement with AI and addressing concerns of equity and ethics head-on, academic communities can better prepare for a future where human and artificial intelligence collaborate to enrich learning at every level.

The insights from Dr. MacNutt and her team not only dispel fears around GenAI misuse but also invite educators to reconsider how they design assignments, evaluate learning outcomes, and support diverse learners in an AI-augmented academic ecosystem. This study offers a hopeful vision that aligns technological progress with educational values, fostering an environment where AI serves as an enabler rather than a disruptor of student success.

Subject of Research: People

Article Title: Reflective writing assignments in the era of GenAI: student behavior and attitudes suggest utility, not futility

News Publication Date: 5-May-2025

Web References:
Advances in Physiology Education Article
DOI Link

References:
MacNutt, M., & Stranges, T. (2025). Reflective writing assignments in the era of GenAI: student behavior and attitudes suggest utility, not futility. Advances in Physiology Education, DOI:10.1152/advan.00241.2024.

Keywords: Education, Education policy, Education technology, Educational assessment, Educational attainment, Educational methods, Learning