Acceptance of generative AI in higher education: A latent profile analysis of policy guidelines

doi:10.21203/rs.3.rs-4515787/v1

Download PDF

Research Article

Acceptance of generative AI in higher education: A latent profile analysis of policy guidelines

https://doi.org/10.21203/rs.3.rs-4515787/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Generative AI tools such as ChatGPT and Bard are quickly changing higher education, bringing both opportunities and challenges. This study examines how top-ranked universities differ in their acceptance of generative AI, applying a latent profile analysis to classify universities based on their acceptance levels and four institutional characteristics: the ratio of international students, citation per faculty, academic reputation, and faculty-student ratio. The results revealed four distinct profiles. Profile 1 includes universities with a strong opposition to unauthorized AI use, underscoring academic integrity, and boasting high international student ratios and research output. Profile 2 consists of universities supportive of responsible AI use, despite lower international presence and research output, highlighting the role of a supportive environment. Profile 3 represents universities with a neutral stance on AI, focusing on ethical usage while having strong international presence but struggling with research output. Profile 4 also adopts a neutral stance, with high academic reputations and research output but moderate international presence and lower faculty-student ratios. These findings are in line with previous research on AI acceptance at the student and faculty levels, highlighting the importance of supportive environments and clear institutional policies. This study provides valuable insights for educators, policymakers, and academic institutions navigating the integration of generative AI technologies.

Educational Psychology

Educational Philosophy and Theory

artificial intelligence

higher education

guidelines

academic integrity

Generative AI tools such as ChatGPT and Bard are quickly changing higher education by offering new opportunities and challenges. Students in the UK, the US, and Japan have rapidly started using these tools, with reports showing high levels of adoption (Masutani, 2023; Nietzel, 2023; Sleator & Hennessey, 2023). For instance, nearly half of the students in the UK, 20% of college students in the US, and 32% of college students in Japan reported using ChatGPT in their studies. The swift adoption of generative AI enhances learning through personalized feedback and assignment aid but risks academic integrity by fostering student reliance. Institutions have responded variably, from bans to full integration. For navigating their use effectively while maintaining academic integrity, clear guidelines are essential. To create effective guidelines, understanding what kind of universities accept generative AI is useful, which is the main research question of the present study.

Benefits and Concerns with Generative AI

A growing body of research has investigated the benefits of generative AI in higher education, such as assisting non-native English speakers, improving art education, aiding idea generation and synthesis, and accurately scoring essays, thereby improving student outcomes (Chan & Hu, 2023; Kasneci et al., 2023; Michel-Villarreal et al., 2023). A mixed-method online survey shows that generative AI assists non-native English-speaking students by aiding idea generation, improving writing structure, and providing personalized feedback (Chan & Lee, 2023). An empirical study finds that a generative AI tool, stable diffusion, enhances art education by allowing the creation of original visual art from natural language descriptions, offering cost-effective possibilities for artistic experimentation and expression (Dehouche & Dehouche, 2023). A commentary piece suggests that generative AI tools aid in idea generation, information synthesis, and summarizing large text datasets (Van Dis et al., 2023). Finally, an empirical study reveals that generative AI effectively accurately scores essays when combined with analysis of linguistic features (Mizumoto & Eguchi, 2023). These studies highlight the potential of generative AI to transform teaching and learning, ultimately improving student outcomes in higher education.

Despite its potential, the rise of generative AI in education presents several challenges, particularly concerning academic integrity. An extensive international survey reveals that generative AI tools including ChatGPT can be misused for plagiarism by enabling students to submit AI-generated essays that seem original, thus compromising the authenticity of their work (Yusuf et al., 2024). A review work suggests that these tools can also create fake references and citations (Kohnke et al., 2023). These misuses threaten academic standards and the originality of student work. Further, excessive dependence on generative AI tools can weaken students’ writing and critical thinking abilities by making them reliant on AI for idea generation and assignment completion. According to a commentary piece, this over-reliance reduces their capacity to think independently and creatively, as they depend on AI-generated content instead of developing their own ideas (Warschauer et al., 2023). An online survey also shows that reliance on generative AI tools hinder the development of essential writing skills, as students may forgo the practice needed to improve their abilities (Chan & Hu, 2023). Furthermore, concerns about generative AI extend to the reputation of higher education institutions as its widespread misuse could undermine trust in the credibility of academic credentials. Several studies find that it is difficult to verify if students truly have the skills and competencies their grades show, which leads to inaccurate assessments of learning outcomes (Dwivedi et al., 2023; Gupta et al., 2024; Ibrahim et al., 2023). This could lead to a decline in the academic reputation of institutions, reducing confidence in the value of their educational credentials. Thus, addressing the misuse of generative AI in education is critical to maintaining academic integrity, fostering real skill development, and protecting the credibility of academic institutions.

Different Guidelines on the Use of Generative AI

Given the detrimental impact of generative AI technologies on conventional evaluation techniques, there is a growing call on higher education institutions to develop extensive policies for implementing these technologies. For instance, a literature review recommends that higher education institutions create clear, easy-to-understand policies detailing the appropriate use of language models in education and outlining the consequences of cheating. In fact, universities worldwide have taken action in response to concerns about generative AI by establishing specific offices or centers and involving university presidents. These entities are tasked with creating guidelines and documents to address ethical issues related to AI. Prominent examples include Stanford University’s Office of Community Standards, Yale University’s Poorvu Center for Teaching and Learning, and the President of Osaka University. In early 2023, over two-thirds of the top 100 universities released guidelines on the use of generative AI technologies.

Recent research has also started to focus on guidelines for generative AI in higher education. A policy analysis study investigates how top-ranked higher education institutions are responding to the rise of generative AI tools in academic assessment practices (Moorhouse et al., 2023). This study selected the top 50 universities from global rankings and collected publicly available guidelines on generative AI use from their official websites to identify common themes and recommendations. The guidelines covered three main areas: academic integrity, assessment design, and communication with students. The guidelines on academic integrity addressed several forms of plagiarism involving generative AI, such as the copying of AI-generated text. Institutions such as the University of California, Berkeley clarified that using AI-generated content without appropriate credit is considered plagiarism. For assessment design, the guidelines proposed rethinking tasks to avoid the misuse of generative AI. This included designing tasks that require critical thinking and incorporating contextual elements. Institutions such as the University of Texas at Austin offered detailed strategies to redesign assessments using AI tools. Communication with students was another crucial area. The guidelines advised instructors to set clear expectations about generative AI use, engage in open discussions about its ethical implications, and collaborate with librarians to educate students on the proper use of these tools.

Another policy analysis study investigates how higher education institutions in the United States have adapted their policies and guidelines in response to the emergence of generative AI tools (McDonald et al., 2024). This study collected documents from 116 US universities classified as high research activity (R1) institutions to understand the guidance for generative AI usage. The results revealed that a majority of universities (63%) encourage the use of generative AI, with many providing detailed guidance for its integration in the classroom. Specifically, 41% of the institutions offer sample syllabi and 50% provide generative AI curriculum and activities. These institutions suggest various classroom activities that involve generative AI, such as brainstorming, writing support, and evaluating the veracity of information. Moreover, around 30% of the universities recommend using generative AI for lesson planning, creating quizzes, and providing personalized feedback to students. Despite this encouragement, some institutions adopt a cautious approach. About 44% of the universities discourage the use of generative AI detection tools due to their unreliability. More than half of the institutions (54%) provide guidance on creating assignments that minimize reliance on generative AI. Privacy concerns are also highlighted by approximately 60% of the institutions, which caution against sharing personal or sensitive data with generative AI. Over half of the institutions (52%) are concerned about ethics, focusing on academic integrity, bias, and intellectual property.

Present Study

The recent policy analysis studies provide valuable insights into how higher education institutions are adapting to the rise of generative AI tools, yet these studies have limitations. They identify general trends in guidelines for generative AI in higher education without deeply exploring the variations in acceptance of generative AI across different types of institutions. Little is known about what kinds of higher education institutions accept generative AI. Therefore, the present study addresses this question with guidelines of generative AI published in 100 top-ranked universities applying latent profile analysis to define and categorize the acceptance of generative AI using four key variables: international student ratio, citation per faculty, academic reputation, and faculty-student ratio.

The methodology of this study entails a quantitative analysis of generative AI guidelines sourced from universities globally. Drawing on previous research (Jobin et al., 2019; Piasecki et al., 2018), we gathered documents that outline principles and guidelines for the use of generative AI in higher education. The study procedure received approval from the research ethics committee at the Center for International Education and Exchange, Osaka University.

Keyword-based Search

A keyword-based web search was conducted on Google.com using a private browsing mode. Before starting the search, all web cookies and browsing history were cleared, and personal accounts were logged out. The search included keywords such as [generative AI guidelines], [generative AI policies], [generative AI statement], and [UNIVERSITY NAME], with the latter being replaced by the names of universities ranked among the top 100 in the Quacquarelli Symonds (QS) university rankings. Results up to the 200th listing were reviewed and screened specifically for documents related to generative AI guidelines. This process yielded 68 unique documents. Moreover, we kept track of relevant literature up to June 30, 2023, to include any newly released documents.

The sample size for this study was determined based on guidelines (Dziak et al., 2014; Tein et al., 2013), which recommend sample sizes for achieving adequate statistical power in latent profile analysis with varying numbers of items and latent classes. For an effect size of 𝑤 = 0.5 (high effect size), the recommended sample sizes are approximately 50–55 for 5 continuous observed variables and 3 latent classes, and 55–60 for 5 variables and 4 latent classes. Given our study design, which included 5 continuous observed variables and the expectation of identifying 3 or 4 latent classes, we aimed for a minimum sample size of 60 to ensure sufficient power for detecting the latent profiles, consistent with achieving 80% power in similar latent profile analysis scenarios with high effect sizes.

Inclusion Criteria

The final collection included target documents such as statements, guidelines, and notices that met the following inclusion criteria: (i) written in English; (ii) published by university-affiliated institutional entities; (iii) explicitly mentioned generative AI or related concepts in their title or description; (iv) articulated a normative ethical stance, indicating a preferred course of action concerning academic integrity.

Assessment of Acceptance of Generative AI

Building on previous research (Ioku et al., 2024), we utilized generative AI, specifically ChatGPT, to evaluate the acceptance of generative AI in higher education through a two-step process. Initially, we defined the acceptance of generative AI as the degree to which an institution endorses or limits the use of generative AI technologies in academic environments. This definition was derived from earlier studies that linked a university’s approach to generative AI with academic integrity (Bin-Nashwan et al., 2023; Cotton et al., 2024; Eke, 2023).

Using this framework, we established four levels of acceptance, which are described in detail in Appendix 1. Level 1 indicates a strong opposition to generative AI, banning its use without permission and treating infractions as misconduct. Level 2 also opposes generative AI, classifying unauthorized use as plagiarism and requiring adherence to ethical standards and permissions. Level 3 takes a neutral stance, neither supporting nor prohibiting generative AI, but emphasizing academic integrity and ethical use with necessary permissions. Level 4 is supportive, promoting the responsible use of generative AI, recognizing its educational benefits, and providing guidelines for its proper application in assessments.

Subsequently, we employed ChatGPT to classify each university’s guidelines into one of these levels of acceptance. For accuracy, ChatGPT evaluated each guideline three times. The reliability of these ratings was high (α = .86). We then averaged these ratings to determine the level of generative AI acceptance.

Measurement in QS University Ranking

We assessed the diversity of every university through the QS University ranking, which has been used in previous studies in higher education (Atici et al., 2021; Aviso et al., 2021; Dobrota et al., 2016; Huang, 2012). The QS Rankings utilized both a reputational survey, which included input from academics and employers, and quantitative indicators, such as citations per faculty, faculty-student ratio, and the proportions of international faculty and students. First, the international student ratio reflects a university’s global presence by comparing the number of international students to the total student population. High ratios in this metric often correlate with a strong academic reputation and research impact, highlighting a university’s international appeal. Second, academic reputation is measured through an international survey, in which academics identify the leading universities in their disciplines. This survey collected responses from over 50,000 academics worldwide over a span of three years, with regional weightings applied to account for differences in response rates. Third, the citation per faculty metric evaluates research productivity. It uses Scopus, a broad research database, to examine the number of citations received by a university’s academic staff over the past five years. Fourth, the employer reputation metric derives from a global survey of more than 20,000 graduate employers, who pinpoint the universities that produce the best graduates, offering a perspective on how institutions are perceived in the job market. Finally, the student-to-faculty ratio indicates the number of academic staff available per student, which is a measure of the institution’s ability to offer small class sizes and personalized attention. It helps assess the university’s teaching quality, as there is no standard international measure for this aspect. We used not only the ratio of international students but also the other indicators to ensure that the diversity effect was distinct from other indicators.

Statistical Analysis

To identify profiles of universities with similar patterns to acceptance of generative AI and four university evaluation indicators, we conducted a latent profile analysis. Latent profile analysis is a statistical technique used to identify subgroups within a population that are internally homogeneous and distinct from each other (Laursen & Hoff, 2006; Wang et al., 2021). To determine the optimal profile model, we compared different models based on information criteria, likelihood ratios, and entropy values (Lee & Chei, 2020; Tein et al., 2013). Lower values of the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) indicate a better-fitting model (Tein et al., 2013). The bootstrap likelihood ratio test (BLRT) is used to compare models with different numbers of profiles, with a significant BLRT p-value (< 0.05) indicating that a model with K profiles is a better fit than one with (K-1) profiles. Higher entropy values (> 0.70) indicate clearer separation between profiles. Additionally, we ensured that the smallest profile had a frequency of at least 5% of the total population to ensure meaningful interpretation (Lee & Chei, 2020).

We performed the latent profile analysis using five variables: acceptance, international students, citation per faculty, academic reputation, and faculty-student ratio. We identified a four-profile model as the optimal solution, as it presented the lowest values for information criteria such as AIC and BIC, a significant BLRT p-value of 0.01, and an acceptable entropy level of 0.88 (Table 1).

Table 1

Fit Indexes for Models with Different Numbers of Profiles
Profile Number	AIC	BIC	Entropy	BLRT p
1	1427.06	1453.11	1
2	1387.05	1428.73	0.89	< 0.01
3	1375.65	1432.97	0.84	< 0.01
4	1361.55	1434.50	0.88	< 0.01
Note. AIC: Akaike information criteria; BIC: Bayesian information criteria; BLRT p: p value of the bootstrap likelihood ratio test

The profiles are differentiated by their characteristic patterns across these variables (Fig. 1). The sample consists of the following proportions: Profile 1 at 29.4%, Profile 2 at 20.6%, Profile 3 at 23.5%, and Profile 4 at 26.5% (Table 2).

[Fig. 1]

Table 2

Universities Included in Each Profile
Profile 1 : 29.4%	Profile 2 : 20.6%	Profile 3 : 23.5%	Profile 4 : 26.5%
Massachusetts Institute of Technology	Peking University	Cornell University	University of California, Berkeley
University of Cambridge	Tsinghua University	University of Toronto	Australian National University
Stanford University	University of Tokyo	Chinese University of Hong Kong	University of Melbourne
University of Oxford	University of Michigan-Ann Arbor	University of British Columbia	The University of Sydney
Harvard University	Northwestern University	Carnegie Mellon University	University of California, Los Angeles
California Institute of Technology	Kyoto University	University of California, San Diego (UCSD)	The University of New South Wales
Imperial College London	Tokyo Institute of Technology	The London School of Economics and Political Science	The University of Queensland
UCL	Brown University	University of Bristol	Monash University
ETH Zurich	Osaka University	The University of Warwick	University of Amsterdam
University of Chicago	Korea University	The Hong Kong Polytechnic University	University of Texas at Austin
University of Pennsylvania	Tohoku University	University of Glasgow	KU Leuven
University of Edinburgh	University of Wisconsin Madison	University of Leeds	National Taiwan University
Princeton University	University of Zurich	KTH Royal Institute of Technology	University of Washington
Yale University	Sungkyunkwan University	University of Birmingham	University of Illinois at Urbana Champaign
Nanyang Technological University, Singapore		Lund University	University of Auckland
University of Hong Kong		Rice University	University of Western Australia
Columbia University			Pennsylvania State University
Johns Hopkins University			Trinity College Dublin, The University of Dublin
New York University
Duke University

Profile 1, which makes up 29.4% of the sample, includes universities that prohibit the unauthorized use of AI by students, viewing it as plagiarism and academic misconduct. At these institutions, students are required to follow ethical guidelines and obtain permission before using AI in their coursework. Universities in Profile 1 have high international student ratios, indicating a strong international presence by comparing the proportion of international students to the overall numbers. Further, these universities also have high scores in citation per faculty and academic reputation, based on research output using Scopus over the most recent five years and a global survey of over 50,000 academics. Furthermore, high faculty-to-student ratios at these universities reflect their capacity to provide individualized attention along with small class sizes.

Profile 2, accounting for 20.6% of the sample, consists of universities that advocate for the responsible and ethical use of generative AI by students. These institutions see AI as a beneficial tool for learning and research and promote its use under strict academic integrity guidelines. Universities in this profile have low international-student ratios, indicating a limited international presence, and also have low citation scores per faculty, suggesting below-average research output. Despite this, they maintain a moderate academic reputation and are known for high student-to-faculty ratios, which allow for personalized supervision and smaller class sizes.

Profile 3, which comprises 23.5% of the sample, consists of universities that adopt a neutral position on generative AI usage. These institutions neither explicitly support nor ban AI tools but stress the significance of academic integrity and ethical AI usage, requiring students to adhere to the guidance given by their course coordinators. Universities in Profile 3 have high international-student ratios, indicating a strong international presence. These universities also struggle with research output, as indicated by low citation per faculty scores, despite having a moderate academic reputation. Moreover, their low faculty-to-student ratios may limit their capacity to deliver individualized support and keep class sizes small.

Profile 4, representing 26.5% of the sample, consists of universities that maintain a neutral stance on generative AI. They stress the importance of academic integrity but do not explicitly promote or ban AI usage. Students are required to follow ethical guidelines and obtain necessary permissions when using AI. These universities have a notable international presence, with above-average ratios of international students and high citation per faculty scores, indicating strong research output. However, they could benefit from enhancing their academic reputation and improving faculty-to-student ratios to attract more international students while upholding high educational standards.

The latent profile analysis of universities’ acceptance of generative AI tools in higher education revealed four distinct profiles, each characterized by unique patterns in acceptance of generative AI, international student ratios, citation per faculty, academic reputation, and faculty-student ratios. These profiles provide valuable insights into how different types of universities are navigating the integration of generative AI, highlighting varying degrees of acceptance and the underlying characteristics associated with their stance.

Profile 1, which comprises 29.4% of the sample, represents universities with a strong opposition to the unauthorized use of generative AI. These institutions emphasize academic integrity and view the unpermitted use of AI as plagiarism. This profile is consistent with previous findings that highlighted concerns about academic integrity and the potential for AI tools to be misused for plagiarism and creating fake references (Kohnke et al., 2023; Yusuf et al., 2024). The high international-student ratios, citation per faculty scores, and academic reputation of these universities suggest that institutions with strong research output and internationalization efforts are more likely to adopt stringent stances on AI usage. The diversity of international student ratios brings a range of opinions and behaviors concerning generative AI, causing uncertainty. Especially, universities with high citation per faculty scores and academic reputations, often prestigious ones, might be likely to perceive risk associated with uncertainty, and thus hesitate to accept the technology.

Profile 2, accounting for 20.6% of the sample, includes universities that are supportive of the responsible and ethical use of generative AI. These institutions recognize the educational value of AI tools and provide clear guidelines for their appropriate use. This support for AI is in line with the benefits identified in the literature, such as assisting non-native English speakers, improving art education, aiding idea generation, and accurately scoring essays (Chan & Hu, 2023; Dehouche & Dehouche, 2023; Mizumoto & Eguchi, 2023). However, these universities have low international-student ratios, citation per faculty scores, and moderate academic reputations, indicating that while they may not be leading in research output or internationalization, they are proactive in integrating innovative technologies probably to be more competitive (Kasneci et al., 2023; Michel-Villarreal et al., 2023). In line with this notion of competitive motivation, profile 2 scores high in faculty-student ratios, which means more academic staff resources are made available to students, such as teaching and supervision.

Profile 3, representing 23.5% of the sample, includes universities with a neutral stance on generative AI. These institutions neither endorse nor prohibit AI use but stress the importance of academic integrity and ethical usage. The high international-student ratios and moderate academic reputations of these universities suggest a strong international presence but a struggle with research output, as indicated by low citation per faculty scores. This profile might indicate that these universities try to be competitive for academic reputation by benefitting from AI tools, but are concerned about the risk associated with uncertainty from diversity (Chan & Lee, 2023).

Profile 4, which accounts for 26.5% of the sample, also represents universities with a neutral stance on generative AI. These institutions have high academic reputations and very high citation per faculty scores, indicating robust research output. They have moderate international presence but very low faculty-student ratios, meaning fewer academic staff resources are available to students for teaching and supervision. The profile suggests that to effectively integrate generative AI and promote student development, universities are expected not to forget to guarantee the quality of education as well as continuing research excellence, which is in line with previous research suggesting that universities manage the affordances and contradictions of AI-generated text in a manner that supports student learning outcomes (Chan & Hu, 2023; Warschauer et al., 2023).

Implications to Practice

To effectively manage the use of generative AI tools like ChatGPT and Bard while ensuring academic integrity, universities are expected to establish clear guidelines. The main research question in the present study focuses on understanding the types of universities that accept generative AI, providing valuable insights for creating these guidelines in three respects.

First, universities may well assess their level of acceptance of generative AI and consider their unique characteristics—such as the ratio of international students, research output, academic reputation, and faculty-student ratios—to tailor their AI integration strategies. Interestingly, institutions with higher international student ratios and stronger research outputs are often less accepting of AI tools, indicating a potential bias in the conversation about AI applications. Rather than relying on assumptions, universities need to base their AI integration strategies on real experiences. By actively engaging with students through surveys and forums, institutions can gain valuable insights into how AI technologies are being used in practice. This approach ensures that policies reflect student needs and expectations. On the other hand, universities with lower international presence and research output might be more inclined to adopt AI tools to improve their competitive edge. For these institutions, creating a supportive environment and establishing clear ethical guidelines for AI use can encourage responsible adoption and maximize the benefits of these technologies.

Second, the different policy orientations towards generative AI, categorized into four groups, are important for faculty members. Students are generally more familiar with information technology, including generative AI than faculty members (Chan & Lee, 2023). The discrepancy of this familiarity between students and faculty members is notable across different regions and educational contexts. In Australia, for instance, universities have been slow to react, often taking countermeasures against AI-related fraud only after incidents have occurred (Slade, 2023). As part of these measures, assessment tasks are being re-evaluated to ensure they accurately measure student learning outcomes, even when AI tools are used. (see an example from University of Texas at Austin). This trend is reflected in recent discussions within our department on what constitutes effective assessment tasks. The practical implications of the findings in the present study can help educators select appropriate university tasks to reference when designing their own assignments.

Third, the different policy orientations towards generative AI can inform governments as well as higher educational institutions. Table 2 shows that almost all Australian universities fall into Profile 4. This may be partly due to government-level guidelines regarding generative AI. Moreover, Australian universities may be less well-known internationally than their US counterparts, which are more likely to be in Profile 1. It is also notable that Profile 2 is composed mainly of East Asian universities. These distinctions highlight the varying levels of AI acceptance and integration policies across different regions, providing further context for developing effective AI strategies in higher education.

Future Directions

Our findings, which delineate profiles of universities based on their stance towards the use of generative AI, can be effectively integrated with previous research on AI acceptance at the student and faculty levels.

A survey study on AI acceptance at the student level found the role of supportive environments and expectancy–value beliefs in fostering students’ intentions to learn AI (Guo & Wang, 2023). Our study supports this by showing that supportive environments at the institutional level can significantly be associated with AI acceptance. Specifically, our Profile 2 universities, which support the responsible and ethical use of AI and provide clear guidelines, reflect the positive impact of a supportive environment. These universities, despite having low international-student ratios and research output, prioritize individualized supervision and ethical guidelines, which is consistent with Wang et al.’s finding on the importance of supportive environments for fostering positive AI adoption intentions among students.

Further, an international survey study on AI acceptance at the student level explored the multifaceted aspects driving the adoption of ChatGPT among university students and highlighted the importance of institutional policies on technology acceptance (Abdaljaleel et al., 2024). Our study extends these insights by providing a typology of institutional responses to AI use. For instance, Profile 1 universities, which are against unauthorized AI use and emphasize ethical guidelines, are in line with Abdaljaleel et al.’s findings that clear institutional policies and guidelines are crucial for responsible AI adoption among students and faculties.

Furthermore, another survey study on AI acceptance at the student and faculty level highlighted the role of habit, performance expectancy, and hedonic motivation in the behavioral intention to adopt AI tools (Strzelecki, 2024). Profiles in our study show how institutional characteristics can shape these factors. For example, Profile 4 universities, with their strong international presence and high research output, might foster a culture where performance expectancy and habitual use of AI tools are more prevalent. This suggests that institutions with high citation per faculty scores and robust international-student ratios can leverage their strengths to promote habitual and motivated use of AI tools, thereby enhancing overall acceptance.

Therefore, future research can build on our findings by examining how specific institutional characteristics and policies influence the individual-level determinants of AI acceptance identified in previous studies. By integrating institutional-level insights with student and faculty-level factors, researchers can develop more holistic strategies for promoting the ethical and effective use of AI in higher education.

Limitations

While the current study has implications to research and practice, it also has some limitations worth noting. The sample used for the study is restricted to guidelines written in English from the top 100 universities worldwide. Although this focus allows for international comparisons, it overlooks universities that provide guidelines solely in their local language, potentially missing important perspectives. Moreover, the majority of top-ranked universities are located in the U.S. and a few other countries, which might have resulted in limited representation from other nations. To address these issues, collaboration with researchers who are native speakers of their respective languages is essential for a more comprehensive understanding. Additionally, concerning the relationship between institutional characteristics and the acceptance of generative AI, the indexes were measured before the development of generative AI guidelines, ensuring the temporal ordering of variables, a crucial component in causal inference. However, this study presents only initial evidence of the relationship, and further validation is necessary to establish a robust link between institutional characteristics and the acceptance of generative AI.

The present study identified profiles of universities based on their acceptance of generative AI and institutional characteristics, highlighting how institutional characteristics are associated with their stance. The latent profile analysis revealed four distinct profiles, each with unique patterns in generative AI acceptance, international student ratios, citation per faculty, academic reputation, and faculty-student ratios. Our findings show that universities with high international student ratios and research output tend to adopt stringent stances on AI usage, emphasizing academic integrity. Conversely, institutions that support responsible and ethical AI use, despite lower research output and international presence, highlight the positive impact of a supportive environment. These insights are consistent with previous research on AI acceptance at the student and faculty levels, underscoring the importance of supportive environments and clear institutional policies. This study provides a foundational understanding of how diverse university policies shape the landscape of generative AI integration.

Conflict of Interest

The authors have no conflict of interests associated with this manuscript.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data Availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Abdaljaleel M, Barakat M, Alsanafi M, Salim NA, Abazid H, Malaeb D, Mohammed AH, Hassan BAR, Wayyes AM, Farhan SS, Khatib S, El, Rahal M, Sahban A, Abdelaziz DH, Mansour NO, AlZayer R, Khalil R, Fekih-Romdhane F, Hallit R, Sallam M (2024) A multinational study on the factors influencing university students’ attitudes and usage of ChatGPT. Scientific Reports, 14, 1983. https://doi.org/10.1038/s41598-024-52549-8
Atici KB, Yasayacak G, Yildiz Y, Ulucan A (2021) Green University and academic performance: An empirical study on UI GreenMetric and World University Rankings. J Clean Prod 291:125289. https://doi.org/https://doi.org/10.1016/j.jclepro.2020.125289
Aviso KB, Demeterio FPA, Janairo JIB, Lucas RIG, Promentilla MAB, Tan RR, Yu DEC (2021) What university attributes predict for graduate employability? Clean Eng Technol 2:100069. https://doi.org/https://doi.org/10.1016/j.clet.2021.100069
Bin-Nashwan SA, Sadallah M, Bouteraa M (2023) Use of ChatGPT in academia: Academic integrity hangs in the balance. Technol Soc 75:102370. https://doi.org/https://doi.org/10.1016/j.techsoc.2023.102370
Chan CKY, Hu W (2023) Students’ voices on generative AI: Perceptions, benefits, and challenges in higher education. Int J Educational Technol High Educ 20(1):43. https://doi.org/10.1186/s41239-023-00411-8
Chan CKY, Lee KKW (2023) The AI generation gap: Are Gen Z students more interested in adopting generative AI such as ChatGPT in teaching and learning than their Gen X and millennial generation teachers? Smart Learn Environ 10:60. https://doi.org/10.1186/s40561-023-00269-3
Cotton DRE, Cotton PA, Shipway JR (2024) Chatting and cheating: Ensuring academic integrity in the era of ChatGPT. Innovations Educ Teach Int 61:228–239. https://doi.org/10.1080/14703297.2023.2190148
Dehouche N, Dehouche K (2023) What’s in a text-to-image prompt? The potential of stable diffusion in visual arts education. Heliyon 9:e16757. https://doi.org/https://doi.org/10.1016/j.heliyon.2023.e16757
Dobrota M, Bulajic M, Bornmann L, Jeremic V (2016) A new approach to the QS university ranking using the composite I-distance indicator: Uncertainty and sensitivity analyses. J Association Inform Sci Technol 67(1):200–211. https://doi.org/https://doi.org/10.1002/asi.23355
Dwivedi YK, Kshetri N, Hughes L, Slade EL, Jeyaraj A, Kar AK, Baabdullah AM, Koohang A, Raghavan V, Ahuja M, Albanna H, Albashrawi MA, Al-Busaidi AS, Balakrishnan J, Barlette Y, Basu S, Bose I, Brooks L, Buhalis D, Wright R (2023) Opinion Paper: So what if ChatGPT wrote it? Multidisciplinary perspectives on opportunities, challenges and implications of generative conversational AI for research, practice and policy. Int J Inf Manag 71:102642. https://doi.org/https://doi.org/10.1016/j.ijinfomgt.2023.102642
Dziak JJ, Lanza ST, Tan X (2014) Effect size, statistical power, and sample size requirements for the bootstrap likelihood ratio test in latent class analysis. Struct Equation Modeling: Multidisciplinary J 21(4):534–552. https://doi.org/10.1080/10705511.2014.919819
Eke DO (2023) ChatGPT and the rise of generative AI: Threat to academic integrity? J Responsible Technol 13:100060. https://doi.org/https://doi.org/10.1016/j.jrt.2023.100060
Guo K, Wang D (2023) To resist it or to embrace it? Examining ChatGPT’s potential to support teacher feedback in EFL writing. Educ Inform Technol. https://doi.org/10.1007/s10639-023-12146-0
Gupta R, Nair K, Mishra M, Ibrahim B, Bhardwaj S (2024) Adoption and impacts of generative artificial intelligence: Theoretical underpinnings and research agenda. Int J Inform Manage Data Insights 4(1):100232. https://doi.org/https://doi.org/10.1016/j.jjimei.2024.100232
Huang M-H (2012) Opening the black box of QS World University Rankings. Res Evaluation 21(1):71–78. https://doi.org/10.1093/reseval/rvr003
Ibrahim H, Liu F, Asim R, Battu B, Benabderrahmane S, Alhafni B, Adnan W, Alhanai T, AlShebli B, Baghdadi R, Bélanger JJ, Beretta E, Celik K, Chaqfeh M, Daqaq MF, Bernoussi Z, El, Fougnie D, Garcia de Soto B, Gandolfi A, Zaki Y (2023) Perception, performance, and detectability of conversational artificial intelligence across 32 university courses. Sci Rep 13(1):12187. https://doi.org/10.1038/s41598-023-38964-3
Ioku T, Kondo S, Watanabe Y (2024) Performance of artificial intelligence: Does artificial intelligence dream of electric sheep. Res Square. https://doi.org/10.21203/rs.3.rs-4469443/v1
Jobin A, Ienca M, Vayena E (2019) The global landscape of AI ethics guidelines. Nat Mach Intell 9:389–399. https://doi.org/10.1038/s42256-019-0088-2
Kasneci E, Sessler K, Küchemann S, Bannert M, Dementieva D, Fischer F, Gasser U, Groh G, Günnemann S, Hüllermeier E, Krusche S, Kutyniok G, Michaeli T, Nerdel C, Pfeffer J, Poquet O, Sailer M, Schmidt A, Seidel T, Kasneci G (2023) ChatGPT for good? On opportunities and challenges of large language models for education. Learn Individual Differences 103:102274. https://doi.org/https://doi.org/10.1016/j.lindif.2023.102274
Kohnke L, Moorhouse BL, Zou D (2023) ChatGPT for language teaching and learning. RELC J 54:537–550. https://doi.org/10.1177/00336882231162868
Laursen B, Hoff E (2006) Person-centered and variable-centered approaches to longitudinal data. Merrill-Palmer Q 52(3):377–389. http://www.jstor.org/stable/23096200
Lee J-Y, Chei MJ (2020) Latent profile analysis of Korean undergraduates’ academic emotions in e-learning environment. Education Tech Research Dev 68(3):1521–1546. https://doi.org/10.1007/s11423-019-09715-x
Masutani F (2023) Survey: 32.4% of college students in Japan say they use ChatGPT. Asashi Shimbun. https://www.asahi.com/ajw/articles/14927968
McDonald N, Johri A, Ali A, Hingle A (2024) Generative artificial intelligence in higher education: Evidence from an analysis of institutional policies and guidelines. ArXiv Preprint ArXiv :240201659
Michel-Villarreal R, Vilalta-Perdomo E, Salinas-Navarro DE, Thierry-Aguilera R, Gerardou FS (2023) Challenges and opportunities of generative AI for higher education as explained by ChatGPT. In Education Sciences (Vol. 13, Issue 9). https://doi.org/10.3390/educsci13090856
Mizumoto A, Eguchi M (2023) Exploring the potential of using an AI language model for automated essay scoring. Res Methods Appl Linguistics 100050. 2https://doi.org/https://doi.org/10.1016/j.rmal.2023.100050
Moorhouse BL, Yeo MA, Wan Y (2023) Generative AI tools and assessment: Guidelines of the world’s top-ranking universities. Computers Educ Open 5:100151. https://doi.org/https://doi.org/10.1016/j.caeo.2023.100151
Nietzel M (2023) More than half of college students believe using ChatGPT to complete assignments is cheating. Forbes. https://www.forbes.com/sites/michaeltnietzel/2023/03/20/more-than-half-of-college-students-believe-using-chatgpt-to-complete-assignments-is-cheating
Piasecki J, Waligora M, Dranseika V (2018) Google search as an additional source in systematic reviews. Sci Eng Ethics 24:809–810. https://doi.org/10.1007/s11948-017-0010-4
Slade L (2023) What is cheating: Students want more clarity on generative AI use at university. 9 News. https://www.9news.com.au/national/chatgpt-university-generative-ai-students-use-technology-to-pass-subjects/12b76e5f-ac79-411c-b443-3fa544a1b783
Sleator L, Hennessey M (2023) Almost half of Cambridge students admit they have used ChatGPT. Times. https://www.thetimes.co.uk/article/cambridge-university-students-chatgpt-ai-degree-2023-rnsv7mw7z
Strzelecki A (2024) Students’ acceptance of ChatGPT in higher education: An extended unified theory of acceptance and use of technology. Innov High Educ 49(2):223–245. https://doi.org/10.1007/s10755-023-09686-1
Tein J-Y, Coxe S, Cham H (2013) Statistical power to detect the correct number of classes in latent profile analysis. Struct Equation Modeling: Multidisciplinary J 20(4):640–657. https://doi.org/10.1080/10705511.2013.824781
Van Dis EAM, Bollen J, Zuidema W, Van Rooij R, Bockting CL (2023) ChatGPT: Five priorities for research. Nature 614(7947):224–226
Wang Y, Shen B, Yu X (2021) A latent profile analysis of EFL learners’ self-efficacy: Associations with academic emotions and language proficiency. System 103:102633. https://doi.org/https://doi.org/10.1016/j.system.2021.102633
Warschauer M, Tseng W, Yim S, Webster T, Jacob S, Du Q, Tate T (2023) The affordances and contradictions of AI-generated text for second language writers. J Second Lang Writ, 62
Yusuf A, Pervin N, Román-González M (2024) Generative AI and the future of higher education: A threat to academic integrity or reformation? Evidence from multicultural perspectives. Int J Educational Technol High Educ 21(1):21. https://doi.org/10.1186/s41239-024-00453-6

Appendix is not available with this version.

The authors declare no competing interests.

Download PDF

Version 1

posted

You are reading this latest preprint version

Acceptance of generative AI in higher education: A latent profile analysis of policy guidelines

Status:

Version 1

Abstract

Figures

Introduction

Benefits and Concerns with Generative AI

Different Guidelines on the Use of Generative AI

Present Study

Method

Inclusion Criteria

Assessment of Acceptance of Generative AI

Measurement in QS University Ranking

Statistical Analysis

Results

Discussion

Implications to Practice

Future Directions

Limitations

Conclusion

Declarations

Conflict of Interest

Funding

Data Availability

References

Appendix

Additional Declarations

Status:

Version 1