Generative artificial intelligence in higher education: an investigation of perceptions, reported use and cultural implications among Caribbean students

Over 100 million people will use generative artificial intelligence (GenAI) to help them complete their work by 2026 (Gartner, 2024). As such, GenAI has been widely acknowledged as a pivotal catalyst for technological advancements in the Fourth Industrial Revolution (Mannuru et al., 2023). Considering the growing debate about the increased use of artificial intelligence (AI) tools, it is important to learn about the perceptions, availability and practical applications of educational AI tools among higher education (HE) students – contributors to the future workforce.

GenAI encompasses a group of machine learning algorithms designed to generate new data samples that mimic existing datasets (Chan & Hu, 2023). It can be used to create new written, visual or audio content, summarize complex data, generate codes, assist with repetitive tasks or make customer service more personalized (Gartner, 2024). Several AI tools primarily cater to learning or educational purposes. Few research studies provide insights into the perceptions and utility of popular GenAI tools, which range from Grammarly GO to ChatGPT, in various parts of the world, with research being especially sparse for developing countries like those in the Caribbean (Adiguzel, Kaya, & Cansu, 2023; Baidoo-Anu & Ansah, 2023; Chan & Hu, 2023).

Although there is a growing interest in the applications and implications of GenAI use globally, AI integration within the Caribbean context presents unique challenges and opportunities. While vibrant enthusiasm exists for these progressive technologies, adaptation rates vary significantly, mainly due to accessibility to institutional resources and technological infrastructure (Camacho-Zuñiga, 2024; UNESCO, 2023). This study goes further by investigating nuanced factors influencing GenAI adoption and implementation in the Caribbean, aligned with cultural diversity, equity, access to resources and data ethics. The findings from this research will add to the literature associated with this unravelling phenomenon within Caribbean spaces and Small Island Developing States in general, where there is a paucity of research (Camacho-Zuñiga, 2024; Mijts, Arens, Buys, & Gielen, 2019).

There is a need for studies on the extent to which tertiary students from various ethnic backgrounds, socioeconomic status and academic areas of study are aware of the types of educational AI tools that exist and their perceptions regarding responsible use of these resources. AI systems often reflect the Western cultural norms of their developers (Hagerty & Rubinov, 2019), which raises concerns about cultural sensitivity and bias, especially in regions like the Caribbean, where diverse cultural and socio-economic factors come into play. Additionally, issues such as limited technological access, content bias and misinformation remain significant barriers to effective AI integration (Chan & Hu, 2023; Harrer, 2023; Maerten & Soydaner, 2023).

Additionally, by incorporating the Biggs 3P theoretical framework, this study aims to provide a novel and comprehensive analysis of how GenAI tools are being used to enhance learning experiences, support educational equity and address the specific challenges faced by Caribbean students in HE. The application of Biggs’ 3P model in this context offers a unique perspective on how GenAI tools can be adapted and utilized to support the specific needs of Caribbean HE students, addressing a critical gap in the existing literature.

AI is a potent data-gathering tool for decision-making (Judkins, Hwang, & Kim, 2024). ChatGPT is a new generation of AI and has become a driver of historical and innovative development (Yu, 2023). ChatGPT is a technology that is transforming how we communicate as a society and is influencing and shaping the production, life and modes of how humans connect (Hill-Yardin, Hutchinson, Laycock, & Spencer, 2023). The advancement of ChatGPT is reshaping humanity and society (Hill-Yardin et al., 2023). It is also having a significant impact within the HE spheres (Kelly, Sullivan, & Strampel, 2023).

Globally, the literature highlights positive attitudes toward using AI technologies, specifically for personalized learning support, writing, brainstorming and other learning assistance strategies (Aldosari, 2020; Aleedy, Atwell, & Meshoul, 2022; Chan & Hu, 2023). GenAI is mainly used in HE to enhance student’s learning experience through its ability to respond to user prompts to generate highly original output (Chan & Hu, 2023). For example, popular reasons for GenAI usage are its text-to-text generators, like ChatGPT, that provide writing assistance to students, especially non-native English-speaking students, in brainstorming ideas and improving writing (Chan & Lee, 2023). Also, text-to-image AI generators, like DALL-E, have been used to teach technical and artistic concepts in areas of arts and design (Dehouche & Dehouche, 2023).

Notwithstanding the utility of GenAI technologies, several limitations and concerns have been raised. A few pressing issues associated with GenAI use in HE include the questions concerning the originality of students’ work, the potential of GenAI tools to stymie improved writing competence, the (un)reliability of AI-generated outputs, breaches of academic integrity and ethics policies, and lack of equal access for students of different backgrounds (Warschauer et al., 2023). Where elements of culture are concerned, GenAI contents may be biased and harmful based on the datasets that power the AI models, as GenAI tools are unable to assess content validity to determine whether the output generated contains misinformation (Chan & Hu, 2023; Harrer, 2023; Maerten & Soydaner, 2023). Furthermore, AI-generated responses to academic writing prompts show that text outputs are mostly original and relevant to the topics but require revisions due to a lack of personal perspectives and reference errors (Kumar, 2023). Constructing appropriate prompts poses a challenge, especially for second-language learners, as certain levels of linguistic skills are still required for effective prompt-generation (Kumar, 2023).

The widespread use of GenAI can pose a challenge to academic integrity in HE, particularly where it concerns the development of holistic and professional competencies (Chan & Hu, 2023; Warschauer et al., 2023). Hence, there is a need for consistent investigation and monitoring as it relates to GenAI use among HE students. By understanding students’ perceptions and concerns, policymakers and educators can develop well-informed guidelines and strategies for responsible and effective implementation of GenAI tools, ultimately enhancing teaching and learning experiences within HE (Chan & Hu, 2023).

Several scholars have noted the impact of GenAI tools, like ChatGPT, on academia (Islam & Islam, 2024). Other studies measured the relationship between students’ trust in GenAI and its impact on their performance in Computational Science courses (Amoozadeh et al., 2023). Trust is crucial in human-AI interactions, especially given humans’ dependence on AI for solving problems (Jacovi, Marasović, Miller, & Goldberg, 2021). Prior experience, familiarity and knowledge of AI technologies enhance trust of, and support for, such technologies (Horowitz, Kahn, Macdonald, & Schneider, 2023; Schepman & Rodway, 2020). Research has shown that persons in technological fields are more prone to support the development of technologies like autonomous vehicles (Bansal, Kockelman, & Singh, 2016; Moody, Bailey, & Zhao, 2020; Payre, Cestac, & Delhomme, 2014) and are more inclined to trust AI technologies given their familiarity and knowledge of them (Horwowitw et al., 2023). Healthcare research has also shown that familiarity with AI positively influences trust (Chandio, Rehman, Bano, Hammed, & Hussain, 2024). One may therefore assume that effect of familiarity on trust in AI technologies is likely to be similar in educational settings.

Despite its potential benefits, there are significant concerns regarding the use of GenAI in academia, like academic honesty and plagiarism (Fuchs, 2023). Students may utilize GenAI tools in the completion of assignments and may submit these works as their own, leading to academic misconduct (Cotton, Cotton, & Shipway, 2024; Dehouche, 2021; Yu, 2023), which may affect their academic reputations. Concerns have been expressed about ChatGPT in particular (Curtis, 2023). Misinformation is another major concern, and some have also cautioned against the use of AI in research settings as it may lead to insufficient data-driven outputs that are biased (Qadir, 2022; Van Dis, Bollen, Zuidema, Van Rooij, & Bockting, 2023). Additionally, some argue that there is little support generated from AI, because users need to provide suitable and relevant input to ensure the responses are both adequate and accurate (Fuchs, 2023).

In the Caribbean, recent studies have focused on the application of AI across several sectors: business and logistics (Foster & Rhoden, 2020), government (Montoya & Rivas, 2019), healthcare (Kitamura, 2023; Martin Saban, Rubinstein, Cejas, & Perez-Acuna, 2023) and climate change mitigation or sustainable development (Salas, Patterson, & de Barros Vidal, 2022). However, studies associated with AI use in education within the Caribbean context are currently very minimal.

Among the limited yet valuable research insights into AI use in education within the Caribbean are studies that provide a deeper understanding of its integration and impacts. Reid (2023) studied perspectives on tech-savvy teaching and the need for technology integration in instructional planning. Camacho-Zuñiga’s (2024) investigated the integration of GenAI into the teaching-learning assessment processes within HE institutions. Myers, Wyss, Villavicencio Peralta, and Coflan (2022) mapped and analysed Digital Learning Platforms in both Latin America and the Caribbean, thus providing technical details about learning platforms. Lastly, Mishra (2019) discussed the usage of data analytics and AI in ensuring quality assurance in the context of HE. These studies suggest great potential in understanding AI and its application within the context of teaching and learning in the Caribbean.

Still, there remains limited research focused on the Caribbean context. Notably, there is an absence of studies examining how students from different academic disciplines, gender groups and ethnic backgrounds engage with these technologies. The ethical concerns, access to resources and potential for bias in AI-generated content further complicate its integration into educational settings. These gaps suggest a need for more focused research that investigates the specific challenges and opportunities of implementing GenAI in the Caribbean HE space, contributing to a deeper understanding of these dynamics.

The integration of GenAI into educational settings has the potential to revolutionize the way students learn and engage with content; however, the risks and challenges such as inequality and integrity cannot be ignored (Liu et al., 2024). This study uses Biggs 3P model (1987) - Presage, Process and Product – as the guiding theoretical framework to explore the implications of GenAI for HE students in the Caribbean.

Biggs 3P model posits that for learning to take place, it is essential to provide learners with a phenomenal experience by providing teaching that centres on fostering change (Sankar & Raju, 2011). The 3P model is important for understanding teaching and learning as a complex and dynamic system that requires collaboration for the enhancement of teachers’ and students’ educational experiences (Freeth & Reeves, 2004). Students’ perception of learning is crucial because their perception impacts their learning environment (Biggs, 1987).

The considerable technological advances faced by the education sector within the past century and the current era of GenAI have called for an adapted approach to teaching that focuses on the 21st-century learner (Cheng & Yim, 2024). Additionally, AI is harnessing tremendous global attention due to its ability to foster innovation (Cui and Wu, 2021); thus, educators should be more intentional about their pedagogical approaches. The three factors within the Biggs 3P model allow for an analysis of how learning environments, teaching strategies, and measurement and evaluation processes can be enhanced in educational settings.

The first component of the 3P model – Presage, pertains to the elements that precede learning (Çelik, 2020). Presage focuses on the context in which learning will occur, influencing how teachers plan their instruction. Another aspect of Presage is learner characteristics (Freeth & Reeves, 2004). When students feel comfortable about their learning environment, which is reflected in the course content, delivery, curriculum, learning resources and assessment methods, this creates positive feelings, leading students to feel confident about learning and their educational journeys (Chan & Hu, 2023). On the other hand, students who doubt their abilities or have negative perceptions of their learning environment may adopt a surface approach to learning, focussing on memorizing facts and meeting minimum requirements (Chan & Hu, 2023). Students tend to express their willingness to use GenAI tools in HE; thus, the focus should be on leveraging these technologies to enhance learning, while implementing policies that can support the appropriate use of GenAI among students (Cheng & Yim, 2024). The right setting must be created for students to feel comfortable about their learning environment. Measures must be put in place to create a conducive learning environment for students to utilize and benefit from GenAI tools.

The second component in the 3P model is the Process factor, which involves the implementation of educational interventions (Freeth & Reeves, 2004). GenAI has become a frequent theme for discussion in the context of teaching and learning. While there are concerns, there are also benefits (Islam & Islam, 2024). Familiarity with technology encourages trust (Horowitz et al., 2023), and many academic students are willing to use GenAI in the learning process (Chan & Hu, 2023). GenAI technologies, despite their limitations, are continuously being developed and refined. Thus, outright rejection of these tools may not be the best approach. Rather, effective systems should be developed to have GenAI effectively interwoven into the teaching and learning process (Yu, 2023).

The last factor in the 3P model is Product, which focuses on course completion, learning outcomes, course satisfaction, perceived usability and transferability (Çelik, 2020). These outcomes may be affected by students’ use of GenAI in the completion of assignments and coursework (Amoozadeh et al., 2023). Product is a significant factor to consider in students’ use of GenAI tools since many of them are willing to use GenAI in the hopes of realizing the benefits which may be derived from its usage (Horowitz et al., 2023). While GenAI usage is prevalent in many developed countries (Chan & Hu, 2023), there is a lack of overall adequate digital infrastructure within the Caribbean (Mont, Del Pozo, Pinto, & del Campo Alcocer, 2020) which suppresses its use within developing territories. Rural areas, in particular, are less likely to have access to technology compared to urban areas in some Caribbean countries (ECLAC, 2019). More concerning is that many Caribbean countries lack reliable internet access, with overall access less than 70% (Mont et al., 2020). For students to fully embrace GenAI, access to technology is a key factor. Ensuring that students have access to both education and the technological benefits gained from GenAI use is essential for achieving positive course outcomes, usability and transferability (Amoozadeh et al., 2023; Çelik, 2020).

A quantitative cross-sectional survey design was applied in the study, and data were gathered from tertiary students from four Caribbean countries. A battery of instruments, consisting primarily of closed-ended items and one open-ended question, was distributed electronically to the sample of students between February and April 2024. A combination of descriptive and inferential statistical tests was used to analyse and summarize data related to key variables of interest, regarding Caribbean tertiary students’ awareness, knowledge, concern and use of GenAI tools.

The target population included students who were 18 years and older, enrolled in various programs across public and private tertiary institutions (colleges, on-site universities, online universities and TVET institutions) in four Caribbean countries (Barbados, Belize, Jamaica and Trinidad and Tobago [T&T]). Tertiary institutions were identified via online platforms, mainly published by the Ministry of Education in each country. The researchers sought permission from relevant personnel attached to each school to gather data from students enrolled at their institutions. Where permission was granted, these individuals were asked to share the survey link with students via their existing communication channels (e.g. emails). The link directed the participants to the survey, which was prefaced by an informed consent page. This was used to notify each participant about the research purpose, including the benefits of advancing educational theory, policy and practice.

The data were gathered through an online survey administered via Qualtrics.com. The survey, adapted from Chan and Hu (2023), comprised closed-ended questions that employed a 5-point Likert scale with responses ranging from “Strongly Agree = 1” to “Strongly Disagree = 5.” One open-ended item was used to capture student responses associated with the participants’ general perceptions of GenAI. The questions captured data surrounding the students’ awareness, willingness to use, concerns and reported use of GenAI tools. Categorical data, aligned with several socio-demographic characteristics, were also captured and used in the analysis. To ascertain whether the respondents understood the instructions and questions in the questionnaire, the survey was first piloted among a small sample of respondents within the target population (n = 30). The feedback obtained from the pilot study suggested overall clarity among the respondents; minor amendments were required.

The quantitative dataset was exported to SPSS V.29 for analysis. Four hundred and ninety-four (n = 494) responses were gathered. One hundred and seventy-two (n = 172) responses were classified as “incomplete” and removed from the dataset. Responses were also deleted due to withdrawal of consent, missing data and completion times that were substantially shorter than the median response time and other issues. This resulted in a final sample of 282, with a 5.8% margin of error at 95% confidence.

Responses for all Likert items were reverse coded as “Strongly Disagree = 1” to “Strongly Agree = 5.” The dataset was subsequently analysed using descriptive and inferential statistical techniques. Demographic data were reported using frequency distributions. The individual scale items, as well as summed variables for each latent construct, were summarized as means and standard deviations. The scales were checked for reliability using the Cronbach alpha estimate of internal consistency, with all scales returning α values > 0.7, indicative of good reliability (awareness [0.836], knowledge [0.881], concerns [0.763], overall concerns [0.897] and willingness to use [0.905]).

The data captured for this study did not meet the assumptions of normality, as indicated by significant Kolmogorov–Smirnov and Shapiro–Wilk tests. Levene’s test for equality of variances also returned a significant statistic, suggesting a violation of the homogeneity assumption. The requirement of equal group sizes was also unmet. Accordingly, non-parametric tests were utilized for inferential analysis of the data: (a) the Spearman’s rank correlation to test for significant associations between the respondents’ perceptions and use of generative AI tools and (b) the Mann–Whitney and Kruskal–Wallis tests to explore group differences for the various perceptual variables in the study. The findings are outlined in Section 4 of this manuscript.

This study was guided by the ethical principles of voluntary participation, informed consent, anonymity and confidentiality. For this study, the protocol was approved by the Institutional Review Board (IRB) of the University of the West Indies Open Campus, Barbados (Application #: Ref: CREC-OC.0180/10/2023) in accordance with “The University of the West Indies Policy and Procedures on Research Ethics”. All participants were informed of the purpose of the research and the level of confidentiality expected from the researcher during the data collection. Proceeding ethical approval of the study, the online survey link was sent to specific institutions. Before commencing the survey, all respondents were required to indicate their willingness to voluntarily participate in the study by clicking “I agree.” Individuals who failed to do so could not view and answer the questionnaire. Thus, voluntary participation was achieved at both the institutional and individual levels. To ensure anonymity and confidentiality, the researchers collected no identifying information from the respondents nor were student records requested from the institutions. The data gathered via the online survey were carefully stored and were only accessible to the researchers of this study.

Table 1 provides a summary of the characteristics of the respondents along several variables. Females dominated the sample at 72.3% (n = 204), while males comprised 27.7% (n = 78) of the sample. The educational backgrounds of the study’s participants, which showed our highest population, emanated from undergraduate students 58% (n = 160) and more studies that pursued non-STEM programs 60.4% (n = 162). Most of the students came from T&T 51.4% (n = 145), followed by Jamaica 32.6% (n = 92) and the last being Barbados 3.5% (n = 10).

As indicated in Table 2, respondents reported a general knowledge of the limitations of GenAI technologies, with means ranging from 3.45 to 3.89. The highest mean score was observed for knowledge of the potential for GenAI technologies to return factually inaccurate information (mean = 3.95, SD = 0.928), while the lowest mean score was observed for the item related to potential biases in GenAI outputs (mean = 3.45, SD = 1.079).

Respondents reported a general willingness to use GenAI technologies, with means ranging from 3.57 to 4.00 (Table 2). The highest mean score was observed for the time-saving benefits of the technologies (mean = 4.00, SD = 0.873), while the lowest mean score was observed for the item referring to the integration of these technologies into future learning practices (mean = 3.57, SD = 1.079).

A total mean score was first calculated for each multi-item scale variable to test whether relationships existed between variables. A Spearman’s rank correlation analysis was then conducted to assess the relationship between perceptions and use of GenAI among participants. As depicted in Table 3, the analysis revealed significant correlations between GenAI use and several perceptual variables. Specifically, moderate, positive correlations were observed with “Awareness” of GenAI technologies (Spearman’s rho, rs(282) = 0.293, p < 0.001) and “Knowledge” of GenAI technologies (rs(282) = 0.201, p < 0.001). Strong, positive correlations were observed with “Willingness to Use” AI technologies (rs(282) = 0.478, p < 0.001) and the perceived educational benefit of AI (rs(281) = 0.437, p < 0.001). No statistically significant correlations were observed between the use of GenAI and concerns about GenAI use nor with overall concerns about the broader issues attributed to AI. This suggests that while concerns exist, they did not significantly deter AI use by study participants.

Nonparametric tests for differences between demographic groups within study location, ethnicity, level of study, sex and STEM variables were conducted on all perception and usage variables (awareness, knowledge, willingness to use, concerns, overall concerns, educational benefit, used GenAI). Kruskal–Wallis tests were used for demographic variables with more than two groups, followed by pairwise comparisons and post-hoc Mann–Whitney U tests for significant results. For demographic variables with only two groups, Mann–Whitney U tests were used. Results yielding significant differences are shown in Table 4.

Respondents identifying as mixed ethnicity (median = 2.89) reported a statistically significant higher level of awareness than those identifying as Black (Afro-Caribbean, African and African Descent) (median = 2.44). Similarly, those enrolled in STEM programs (median = 2.78) reported a higher level of awareness than those in non-STEM programs (median = 2.44). Statistically significant differences were observed in knowledge between postgraduate students (median = 4.00) and undergraduate students (median = 3.75) and between postgraduate students and those studying in non-degree programs (median = 3.50). A statistically significant difference was also observed in knowledge between males (median = 3.83) and females (median = 3.58). Males also reported a higher willingness to use (median = 3.94) than females (median = 3.75). Significant differences were observed in overall concerns between respondents studying in Jamaica (median = 3.33) versus those studying in T&T (median = 4.00). All other between-group comparisons returned non-significant results.

Our findings revealed that there was a higher knowledge level of GenAI among students enrolled in postgraduate programs versus undergraduate and non-degree programs, and there was a higher level of awareness of GenAI among STEM students compared to their non-STEM counterparts. These findings indicate that students who are pursuing higher degrees in STEM fields have a relatively higher openness and readiness to learn about GenAI applications. “Openness and readiness to learn” are aligned with Biggs’ Presage factor – learner characteristics (Freeth & Reeves, 2004), which represent a vital foundation for technological adaptation and wider learning.

The highest mean scores were observed for the time-saving benefits of the technologies and the perceived benefits of using GenAI technologies for learning purposes. There was also a strong positive correlation observed between willingness to use and the perceived educational benefit of GenAI. These results reflect a strong inclination among students toward the practical application of GenAI tools as reflected in the literature (Adeshola & Adepoju, 2023; Bahaw, Forgenie, Sadiq, & Sookhai, 2025; Chan & Hu, 2023; Chan and Hu, 2023). Halaweh (2023) argues that educators should allow the use of GenAI, as students will be using it regardless of educator instruction. Interestingly, our results reveal a more widespread use of GenAI among students with advanced knowledge of its limitations. This suggests that students who understand the limitations of GenAI tools may be better equipped to use them effectively.

Generative AI-powered tools can also be used to facilitate collaborative projects through student networks across institutions in the Caribbean and beyond. HE institutions in the Caribbean can provide explicit guidelines that govern students’ use of GenAI technologies to promote learning (Adeshola & Adepoju, 2023). Given the idea that students engage with GenAI tools, for example as learning assistants, educators can engage in open dialogue with students around appropriate uses of GenAI, including the importance of academic honesty, and provide them with knowledge of its limitations, such as biased data, non-current information and its potential to generate incorrect outputs (Lo, 2023). When students understand the limitations of GenAI, they can take advantage of its strengths while minimizing the effects of its weaknesses on their learning processes and outcomes.

GenAI perceptions were not uniform across different educational levels, as postgraduate students had higher knowledge about such limitations when compared with students enrolled in undergraduate and non-degree programs. This superior level of knowledge of the limitations of GenAI technologies among postgraduate students suggests that they may have more opportunities to use it, as increased opportunities to use GenAI tools such as ChatGPT may lead to discovery of its weaknesses through hands-on use (Fütterer et al., 2023).

We found a higher level of awareness of GenAI among STEM students compared to their non-STEM counterparts. While STEM students may be more aware, this does not necessarily imply a comprehensive understanding of GenAI, its capabilities or its limitations, as there was no significant difference in knowledge of GenAI limitations or usage rates. The difference in awareness might be a result of curricular differences. STEM curricula are more likely to include elements related to technology, including AI. This exposure, however minimal, could contribute to a higher level of awareness among STEM as opposed to non-STEM students, who might have curricula focused on completely different areas. As GenAI technologies such as ChatGPT pose threats of cheating on online exams, essay generation and diminished critical thinking (Rahman & Watanobe, 2023), one can take on a more cynical perspective. As 50% of STEM entrant students leave STEM fields (Chen, 2009, 2013; Romash, 2019), STEM students may find academic dishonesty more desirable. A study by Roy and Edwards (2023) found that the top motivation for STEM students to cheat was time pressure or a heavy workload. Given this study’s findings that the highest mean score was observed for the time-saving benefits of the technologies, future studies can further investigate the social pressure and time-saving benefits of GenAI technologies.

This study unearthed an interesting finding with ethnic groups’ responses to GenAI. Persons who identify as mixed ethnicity possess higher levels of awareness than those students who identify as Afro-Caribbean, African and African Descent. This finding mirrors past research, which showed that students from diverse ethnic groups had different learning responses toward GenAI as they came from various cultural backgrounds (Tsai, Ma, Chang, & Lai, 2022). The Caribbean is culturally diverse; the social, cultural and ethnic diversity of each Caribbean country is influenced by their colonial history and its influential role in the social, economic and political development in the region (Stephenson, 2018). Such background impacts various choices among Caribbean students, for example concerning academia. Hence, it is important to consider the unique needs of people with various ethnic backgrounds. Findings about cultural diversity are important for crafting effective policy responses in the Caribbean. An understanding of student diversity can help educational practitioners enact policy decisions associated with GenAI application. Ideas for specific educational initiatives can be generated to help students gain greater awareness of the benefits and challenges of GenAI, which can foster more responsible use of these technologies.

The application of the 3P model within the Caribbean context allows for a critical evaluation of the challenges and opportunities in GenAI integration and adoption in the Caribbean region. As it relates to Presage, the main challenges in the region include geographic and infrastructural barriers (Mont et al., 2020; ELAC, 2019). Quite often, technological infrastructure and internet access are challenging not only for educational institutions but also for students who live in rural areas. Innovative strategies and financial networking are required to allow for available and accessible infrastructure that can be used to create or enhance interactive virtual learning spaces, particularly for students residing in remote areas.

Another barrier that can be associated with the Presage domain is cultural diversity. Countries like Jamaica and T&T comprise many culturally descended and mixed groups, such as African, East Indian, Asian, European, Hispanic and indigenous groups, such as the Taino. Factors such as a lack of unity among the Caribbean multiculturality can perpetuate barriers associated with cultural diversity and responsible AI adaptation. To achieve equal opportunity, GenAI tools and content need to be culturally sensitive, relevant and inclusive to cater to the needs of multicultural students.

Challenges related to Process include the continued use of traditional teaching methods that are not adapted to tech-savvy students with an appetite for digital and interactive content. Thus, educational interventions must be progressively student-centred. Instructional processes must be conducive to learning, thereby supporting students’ success and enabling educational institutions to contribute to human capital. This is vital in harnessing professional capacity across the region, increasing the number of technologically savvy citizens and having a digital presence on the world stage. When GenAI is integrated ethically, it can support adaptive learning systems that adjust content and assessments based on student progress and performance (UNESCO, 2023). This challenge is perpetuated by the degree of readiness and willingness of in-service and pre-service professional development for effective integration of AI tools.

Under the Product domain, educational testing and evaluation processes require careful guidance through policies and ethical guidance. This is mainly to ensure the overall integrity of the institutional assessment process. Added to this, ensuring equity is important at all levels, such as technology access and ethical use. The purpose of this is to reduce barriers to equal opportunity, which in turn will reduce disparities in educational outcomes. While students are willing to use GenAI, access remains challenging in the Caribbean Region, especially among the rural population (ELAC, 2019). Thus, policymakers, educational leaders and administrators must be more sensitive to the needs of students and provide them with relevant access to technology in their academic journey that will provide them with the tools to succeed.

Providing access is imperative as it can help with the equitable distribution of resources regardless of gender, race and socioeconomic background such that no student can be left behind. This is especially essential given our results which indicate that students of Black ethnicity had lower levels of awareness when compared to persons of mixed race. This is paramount because, in the Caribbean, the black population is usually marginalized, and the stereotypes this population encounters already result in serious consequences that other ethnic groups do not face (Ryan, Rampersad, Bernard, Mohammed, & Thorpe, 2013). Additionally, Black students are subject to severe hardships within the school system when compared to other ethnic groups (Ufoegbune, 2017). Thus, equity is key in providing access to, and adoption of, modern technology. More must be done to identify and reduce racial, gender or socioeconomic barriers associated with technological (GenAI) adoption.

On a broader (global) scale, findings indicate similar experiences and socioeconomic barriers among marginalized, minority groups, such as Asian, African-descended and Indigenous tribal groups in developed countries like the United States of America, Canada and Europe. Similar trends related to cultural representation, cultural sensitivity, content bias and associated data ethics challenges exist worldwide (Camacho-Zuñiga, 2024; Chan & Hu, 2023; Liu et al., 2024; Qadir, 2022; Van Dis et al., 2023). However, one must also remember that developed countries have relatively more resources, which provides students with better access to technology enabling them to utilize AI technologies in comparison to developing countries.

The adoption and inclusion of GenAI tools is causing a rapid transformation in education, with Halaweh (2023) suggesting that applications such as ChatGPT will foster creativity and innovation through more involved assessments such as reflection notes and presentations. GenAI tools such as ChatGPT should be used in education in a way that complements and aids human teaching and learning (Božić, 2023). To properly use these tools, students around the world must be able to navigate and interface with GenAI tools and understand its constraints, as well as the social, moral and emotional aspects of using it (Adeshola & Adepoju, 2023). Therefore, proper use of GenAI tools may be overlooked by those with little digital competency (Adeshola & Adepoju, 2023). Despite differences in knowledge about the limitations of GenAI, students’ attitudes and behaviours towards AI are relatively consistent across different levels of study. This uniformity could be due to a common underlying cultural or societal perspective on GenAI that similarly influences all students.

This study investigated students’ awareness, familiarity, willingness to engage, perceived benefits and challenges and the utility of educational GenAI applications across four Caribbean countries: Barbados, Belize, Jamaica and T&T. Our findings revealed statistically significant correlations between awareness, knowledge and willingness to use GenAI and its perceived educational benefits. Notably, no correlation was found between concerns about GenAI and its use, indicating that concerns may not significantly hinder adoption. Differences in awareness between mixed ethnicity students and Afro-Caribbean, African and African Descent students highlight the need for culturally sensitive approaches to promoting equitable access.

This study also has three major limitations. First, though the use of Likert items allowed us to capture, to some extent, initial individual perceptions, our predominant use of closed-ended questions prevented us from exploring in-depth perspectives about the use of GenAI AI tools in educational settings.

Second, since the sample for this study comprised post-secondary students in four Caribbean countries, a possible limitation, given the modest sample size and the use of convenience sampling, is that the results obtained in this study may not be generalizable to the broader Caribbean context. The results may also prove inapplicable in other (non-Caribbean). Finally, the sample was skewed according to sex, with a 72% female majority. Thus, while the study aims to establish a foundational understanding of the topic, this understanding can be further built upon in future research.

Our study contributes to the field by offering baseline data on GenAI awareness and usage in an under-researched region, providing valuable empirical evidence for the Caribbean context. This study also provides theoretical contributions by contextualizing GenAI adoption using the Biggs 3P framework and offering practical insights for supporting diverse educational needs. The findings emphasize the need to devise educational policies that promote ethical GenAI use, to consider cultural diversity and to ensure equitable access. Educators should foster open discussions with students about the strengths and limitations of GenAI to enhance learning outcomes and responsible usage. Future research should explore these findings with larger and more diverse samples to enhance generalizability and deepen understanding of cultural influences on GenAI adoption.

RMM: Conceptualization, data curation, project administration, introduction and conclusion; AM: data curation and literature review; AS: data curation and methodology; MM: discussion; BM: software, formal analysis, validation, investigation. All authors: writing – review and editing and approved the final manuscript.

For this study, the protocol was approved by the Institutional Review Board (IRB) of the University of the West Indies Open Campus, Barbados (Application no.: Ref: CREC-OC.0180/10/2023) in accordance with “The University of the West Indies Policy and Procedures on Research Ethics”. All participants were informed of the purpose of the research and the level of confidentially expected from the researcher during the data collection.

Informed consent was gained from all survey respondents prior to their participation in the study. The participants were all adults who freely and willingly consented to complete the survey. The instrument used in this study was prefaced by an informed consent letter, which stated: “Your participation in this survey is voluntary. You may refuse to take part in the research or exit the survey at any time without penalty. You are free to decline to answer any question you do not wish to answer”.