2.1 Migrants and refugees and digital literacies

Many adults who are EAL learners have come to settle in a new country. Settlement in a new country is both a short-term and a long-term process that requires migrants and refugees to access, understand and trust complex information about how to live, learn, work and socialise in their new contexts (Borkert, Fisher & Yafi, Reference Borkert, Fisher and Yafi2018; Potocky, Reference Potocky2021). Migrants, who choose to move to a new country, and refugees, who may be forcibly displaced and are unable to return to their home country safely, must identify strategies for navigating and processing information while on the move and once they have settled into their new country. However, migrants and refugees are often more hesitant to trust online information or government authorities based on their previous lived experiences (Lloyd, Reference Lloyd2020; Safarov, Reference Safarov2021). Migrants and refugees often face “information precarity”, a term that captures the experiences of lack of access to information during their journey to their new country and therefore are vulnerable to misinformation (Şanlıer Yüksel, Reference Şanlıer Yüksel2022).

Despite this precarity, which only furthers feelings of information mistrust, digital technologies, particularly the use of smartphones, have been used as a way for refugees to stay connected and validate information through connected diasporic and transnational networks (Dekker, Engbersen, Klaver & Vonk, Reference Dekker, Engbersen, Klaver and Vonk2018; Gillespie, Osseiran & Cheesman, Reference Gillespie, Osseiran and Cheesman2018; Nedelcu & Soysüren, Reference Nedelcu and Soysüren2022). For example, Şanlıer Yüksel (Reference Şanlıer Yüksel2022) found that migrants, asylum seekers and refugees in Turkey utilised and navigated digital spaces, particularly using smartphones, not only to maintain their diasporic and transnational connections but also to interact and connect with their local contexts while on the move and once settled. This example highlights the knowledge and skills that migrants and refugees must possess so they can gain access to and understanding of key information while also effectively using, appraising and communicating the information they receive in digital spaces (Ndofor-Tah et al., Reference Ndofor-Tah, Strang, Phillimore, Morrice, Michael, Wood and Simmons2019; Potocky, Reference Potocky2021).

Digital literacies are increasingly recognised as an essential competency for all members of contemporary society, including migrants and refugees (Ndofor-Tah et al., Reference Ndofor-Tah, Strang, Phillimore, Morrice, Michael, Wood and Simmons2019; Potocky, Reference Potocky2021). Digital literacies have the capacity to empower refugees to obtain and evaluate information during their journey and settlement and become agential in how they employ digital technologies (Nedelcu & Soysüren, Reference Nedelcu and Soysüren2022; Safarov, Reference Safarov2021). Indeed, digital technologies can alleviate the challenges faced by migrants and refugees by mediating access to information and facilitating communication, ultimately supporting their wellbeing (Ekoh, Okolie, Nnadi, Oyinlola & Walsh, Reference Ekoh, Okolie, Nnadi, Oyinlola and Walsh2023). Digital empowerment allows migrants and refugees to actively decide how they use digital technologies to obtain information, overcome challenges, and avoid misinformation and/or digital surveillance (Nedelcu & Soysüren, Reference Nedelcu and Soysüren2022; Safarov, Reference Safarov2021). There have been calls for a more comprehensive and holistic instructional paradigm for adult EAL learners with migrant and refugee backgrounds that enables full participation in society (McHugh & Doxsee, Reference McHugh and Doxsee2018), with digital literacies as key to their long-term outcomes after settlement (Pegrum, Hockly & Dudeney, Reference Pegrum, Hockly and Dudeney2022).

2.2 AI literacies

Given the widespread adoption of AI for surveillance by governments and governance authorities (Kaneti, Reference Kaneti2023; Nedelcu & Soysüren, Reference Nedelcu and Soysüren2022) and its increasing utilisation in both higher education and general education schooling contexts (Casal-Otero et al., Reference Casal-Otero, Catala, Fernández-Morante, Taboada, Cebreiro and Barro2023; Chen, Tallant & Selig, Reference Chen, Tallant and Selig2024; Pretorius, Reference Pretorius2023), the development of digital literacies among migrants and refugees must also include a focus on AI literacies. Drawing upon the concept of digital literacies, we define AI literacy as knowing how to utilise and interact with AI technologies (including generative AI), while also being able to evaluate their trustworthiness and consider the ethics of their use (Chen et al., Reference Chen, Tallant and Selig2024; Pretorius, Reference Pretorius2023).

Several studies have recently emerged exploring how AI literacy principles can be developed in both higher education (Chen et al., Reference Chen, Tallant and Selig2024; Pretorius, Reference Pretorius2023) and schooling contexts (Casal-Otero et al., Reference Casal-Otero, Catala, Fernández-Morante, Taboada, Cebreiro and Barro2023). In higher education contexts, students often utilise generative AI to prompt and generate ideas for assignments (Chen et al., Reference Chen, Tallant and Selig2024), aiding in the process of completing these assignments (Pretorius, Reference Pretorius2023). A systematic review by Casal-Otero et al. (Reference Casal-Otero, Catala, Fernández-Morante, Taboada, Cebreiro and Barro2023) revealed limited studies that have assessed student knowledge after interactions with and/or learning about AI.

Despite increasing attention to the development of AI literacy, there has been limited empirical evidence of developing AI literacy in learning contexts that involve EAL learners who may have a migrant and refugee background, and there is also a lack of understanding of how educators can specifically support the development of AI literacy. Furthermore, despite a focus on identifying the potential and limitations of generative AI in language learning (Huang, Zou, Cheng, Chen & Xie, Reference Huang, Zou, Cheng, Chen and Xie2023; Muñoz-Basols, Neville, Lafford & Godev, Reference Muñoz-Basols, Neville, Lafford and Godev2023), there is limited empirical evidence of how students develop AI literacies as part of language learning and in other educational programs, particularly in the context of the needs of migrants and refugees, post-settlement.

4.1 Research design

This research employs data from a larger mixed-methods project, of which only part is reported for this article. Our approach is qualitative, using some quantitative descriptive statistics to support crystallisation (Ellingson, Reference Ellingson2009). Multiple case studies were employed to explore the knowledge and attitudes of leaders, educators, and students regarding implementing generative AI within learning programs for adult EAL learners. The use of multiple case studies facilitated a comprehensive exploration, offering a broader and more detailed understanding within and across individual cases (Gustafsson, Reference Gustafsson2017). Three key stakeholder groups were involved: educational leaders within an educational institutional context, educators, and students who identified as having backgrounds where English is another language. Each group was treated as a singular case. The project was conducted in most states across Australia (Table 3), with participants drawn from a range of training providers. Information about the visa type (e.g. refugee or migrant) of adult EAL learners was not available, but the language background implied that these adults entered Australia through a range of visa pathways.

Table 2. Survey details

Table 3. Demographic details of the survey participants

The primary objective of conducting individual case studies was to enhance understanding of the diversity inherent within each group and then focus on examining variations in knowledge and attitudes between cases across different contextual settings (Gustafsson, Reference Gustafsson2017).

4.2 Data collection

A mixed-methods research design was adopted that employed a multiple case study methodology with quantitative and qualitative components to understand attitudes to and uses of generative AI. Using multiple case studies allowed the research team to compare different participant groups, which counted as cases in this research: educational leaders, teachers, and students. Data collection involved surveys and focus group sessions with each case/participant group. Seeking to generate both quantitative and qualitative data, we structured the survey instrument to explore four distinct categories, as presented in Table 2. Table 2 also includes some examples of the survey items from the larger survey.

The Qualtrics platform was used to facilitate a national survey, representing participants across Australia. Demographic details of the survey participants are provided in Table 3.

Table 3 indicates that most leaders, teachers, and students are in an older age bracket, and the majority are female. For the student cohort, the large majority do not speak English at home and are likely to come from migrant and refugee backgrounds, and most live in urban areas.

After completing the survey, the respondents within each group were invited to participate in focus groups. Table 4 provides details of the focus groups. Focus groups aimed to facilitate open-ended discussions exploring the patterns and themes from the survey responses and nuances and ambiguities. All focus groups were conducted online via Zoom and involved participants from various states and program providers.

Table 4. Focus groups details

4.3 Data analysis

Qualitative data analysis employed a deductive approach, leveraging the four perceptual categories presented in Table 1. These categories were adopted as themes in the process of coding the data. Coding using the perceptual categories was undertaken by two of the researchers. Each researcher coded one selected focus group separately and then met to compare the coding outcome to establish intercoder reliability. A high level of intercoder reliability was observed, being above 90%. Coding of all the focus groups was then completed by three researchers.

Data for leaders, educators, and students were analysed separately within each category. This disaggregated analysis, while limiting the cohesive reporting of what occurred within each focus group, facilitated a systematic understanding of each participant group’s specific perspectives and understandings regarding the use of generative AI within the program and enabled cross-comparison of data.

5.1 Functional knowledge or understanding (FK)

The survey data from student participants indicate that over one third of adult EAL participants were aware of generative AI (36%; n = 63), one-third knew how it worked (30%; n = 52), and even fewer have utilised it (22%; n = 39), suggesting a need for increased educational support in this area.

The study found that the leaders and teachers had significant concerns about students’ capacity to understand how generative AI works for practical applications and real-life situations. This is demonstrated by qualitative survey data indicating concerns about “plagiarism” and “over-reliance” as significant challenges alongside the students’ capacity to evaluate AI-generated content. This concern was further discussed in the focus group of leaders:

Leaders perceived students as lacking the necessary critical understandings to evaluate the credibility and accuracy of the information and make ethical decisions when using AI-generated content for learning purposes. The teachers were also concerned about students not using AI in a critical and ethical way: “They’re going to be using AI to fool their lecturers” (LFG).

However, their larger concern was about students’ capacity to understand the nature of the technology and utilising it at an operational level:

The data reflect teachers’ views that generative AI is difficult to understand, as it is a complex technology that offers capabilities that we have not seen before. They imply, in quite a deficit language, that this new technology is especially difficult for CALD learners due to the language barrier. The data also suggest that teachers have some understanding of the complexities involved in prompt engineering, both as a starting point of interaction with generative AI and as a critical aspect of successful interaction for quality output. However, they believed that students would not be able to understand how to craft a suitable prompt, as it is “well above” the students’ capacity. The teachers’ apparent deficit language and their presumptions about learner capabilities in this context suggest they may be undervaluing their students’ capacities for higher-order thinking and critical engagement in home languages, including their capacities for translanguaging using dictation and translation tools that utilise generative AI.

Survey data indicate a basic understanding of generative AI among some students. Nearly four in 10 (39%; n = 23) survey respondents who have used generative AI (30%; n = 29) reported a general understanding of how it works. Focus group discussions further revealed that participants demonstrated some functional knowledge of AI, with one stating, “I have used ChatGPT for quite some time [for] basic information and inquiry … But I know it is AI. There is no person in the background” (SFG). Participants also exhibited awareness of both general-purpose AI tools like ChatGPT and specialised applications tailored to specific needs. As one student noted, “ChatGPT is kind of a general AI, but you have […] many specific ones” (SFG). This highlights some adaptive engagement with technology and understanding of its functionality. However, most student participants had low levels of understanding of what AI is and its potential uses.

5.2 Perceptions about ease of use (EU)

Reflecting on the perceptions about students’ understanding of AI, many participating leaders thought that it would not be easy for EAL students to learn and use generative AI:

At the same time, leaders strongly believed in the potential of generative AI for students, and this corresponds to the understanding exhibited by some students. Survey data support this, with 60% (n = 28) agreeing it is crucial for future employability and 62% (n = 29) seeing its benefits for students. However, as leaders indicated in the focus group, poor digital access and literacy among students pose significant barriers to harnessing AI’s potential.

Teachers in the survey shared a similar concern, also noting that language proficiency makes it difficult for students to use generative AI easily, particularly those at the very beginning of their English language learning journey:

Interestingly, teachers’ opinions in the focus groups about students’ capacity to use generative AI varied:

The quotes from the teacher participants in this study emphasise the importance of prior digital literacy for EAL students in learning programs. The teachers position English language proficiency and digital literacy as prerequisite skills for engagement with generative AI rather than as potential outcomes of engagement. The first quote suggests that students may not have uniform skill levels, inferring that those with a background in digital literacy might find it easier to adapt to and efficiently use new technologies like generative AI. The second quote is quite upbeat about AI and learners’ capacity and motivation to learn efficiently, which could make them receptive to integrating AI into their digital literacy curriculum. This suggests that introducing AI at higher levels could enhance students’ digital literacy skills, potentially making it easier for them to use this technology effectively.

Some participants in the three focus groups conducted with students expressed a positive outlook for generative AI in their learning and lives, noting its efficiency and utility in language learning. Indeed, one participant stated that it enabled them “to make [my] English stronger and stronger by using ChatGPT (SFG)”. However, users are aware of the technology’s limitations, acknowledging occasional inaccuracies and the need to verify AI-generated content, which might suggest that it is also complex to use. Concerns over potential biases in AI responses also surface in the focus group discussion, emphasising the importance of critical engagement with the technology.

Despite this, there appears to be a strong appetite for further exposure to AI to “enhance our daily life” (SFG), indicating recognition of its potential to streamline tasks and improve productivity, making life easier. For adult learners, AI offers practical applications in navigating the often complex and bureaucratic process of establishing a life in Australia. Beyond simple document searches, tools like ChatGPT can provide step-by-step guidance for applying for government services, such as detailing the necessary forms, eligibility criteria, and submission procedures, with one participant appreciating how AI clarifies “how to find a government document […] very helpful” (SFG).

Collectively, these reflections in the student focus groups suggest that although generative AI is valued for its practical benefits and ease of use for some specific tasks, users also show some critical perspective on its use and its inherent complexities.

5.3 Usefulness or relevance of the technology (UR)

Survey results reveal a reasonable acceptance among students regarding generative AI’s role in job acquisition and future employment, with approximately half of the students positive about its practical benefits. Fifty percent (n = 47) agreed that “if I know generative AI, it will be easier to get a job”. Similarly, 51% (n = 48) agreed that “generative AI can help me in my future job”.

When considering the usefulness or relevance of generative AI to students’ lives, the majority of leaders participating in the survey agreed that students need to know how to use generative AI for their future employability (60%; n = 28) and that generative AI can be beneficial in their program (62%; n = 29), suggesting a positive outlook and relevance for students. One leader stated in the survey, “Students using ChatGPT can access information in their own language or have access to texts in English that are useful and related to their own needs”. The focus here is on access to text rather than usefulness based on the capacity to use the technology in practical ways. However, the leaders in the focus group went beyond the issue of access to seeing ChatGPT as important for both understanding texts and expanding what they can do:

The teachers recognised the potential of technology to assist students in their daily lives, particularly in overcoming challenges with complex text production or interpretation. They viewed it as a supportive tool that can enhance students’ abilities, but how this would be integrated into education to facilitate learning and skill expansion is less certain.

At the same time, there was much caution regarding the relevance and usefulness of generative AI for the student:

The leaders participating in the focus group thought that factors such as individual learning abilities, vocational aspirations, and personal circumstances play an important role in determining whether generative AI is useful for people. However, they generally acknowledged their emerging knowledge about AI, noting their desire to learn more about and understand how this new technology works and how it can be useful for their cohorts of students.

Similar trends were evident in the teachers’ survey data: 63% (n = 28) agreed that students need to know how to use generative AI for their future employability and 55% (n = 21) agreed that generative AI can benefit students in the program. Two participants in the survey noted the following:

However, there are also alternative opinions and clear hesitation:

The quote implies that the teacher believed that students did not know much about generative AI and thus could not envision its potential application and usefulness for their lives. The participant also noted a lack of interest in and perceived value of generative AI, attributing this to it being “a novelty” and students’ poor digital literacies. This set of quotes reveals the critical lack of integration of generative AI into the educational settings and perhaps awareness of the challenges that the students face after settlement.

From the qualitative data of the student participants in the focus groups, it is evident that generative AI is a highly relevant and useful technology. One participant describes the AI’s utility in streamlining interactions with essential government services online, noting, “it’s going to take you three days to answer the question. So, if you have something like that, I mean, that’s very useful” (SFG). This quote suggests a significant reduction in the time spent on tasks pivotal to surviving and prospering in Australia. Another finds value in language learning, “looking for books, and texts and things to learn English” (SGF), pointing to the AI’s role as an educational aid.

Furthermore, users readily appreciated the extensive, detailed information across multiple languages through generative AI, enhancing their ability to access and understand a broader range of content (SFG). According to several student participants, AI also serves as a homework assistant, especially for those managing multilingual households (SFG). For one participant oriented to becoming an entrepreneur, it supports business marketing efforts, helping “to create images for special promotions” (SFG), which indicates its impact on economic activities. Collectively, these reflections stress the potential usefulness of generative AI to support integration and empowerment.

5.4 Intention to apply the technology (IA)

The survey data (SS) indicate cautious attitudes among students toward future intention to use generative AI, with only 20% (n = 18) planning to use it and 16% (n = 15) believing in its universal adoption. By contrast, when asked about the likelihood of implementing generative AI in learning programs, leaders seem to have a high intention to use it in the future (62%; n = 29). However, most of their examples were related to their leadership and administrative duties to speed up and streamline their activities rather than considering pedagogical applications in the classroom to equip learners with knowledge and skills for generative AI. In this sense, their approach to implementing AI is seen as utilitarian rather than visionary.

In addition, many leaders in the survey felt uncertain about how generative AI can be implemented, citing their nascent understanding of this technology at the time of the study:

Nevertheless, most of the participants (92%; n = 42) expressed interest in learning more about generative AI, though there was a tendency to describe this technology as relevant sometime in the future, so no actual implementation plan was envisioned at this stage.

Similarly, teachers strongly intended to use generative AI in the future (66%; n = 25) and learn about it (87%; n = 33). There was a lot of interest in using generative AI for lesson planning, developing resources, and different administration tasks, or what might be seen as instrumental outcomes from the technology rather than innovative pedagogical possibilities and creative outputs. There were limited responses in the data set, indicating interest in using generative AI with students in class:

At the same time, there were 18 open-ended survey responses suggesting a very strong stance that AI should not be used in the program, especially with beginning learners:

These excerpts from the data suggest clear scepticism and apprehension towards implementing generative AI in classrooms. Teachers expressed concerns about its potential negative impact on their students including those learning EAL. Additionally, doubts arise regarding its accuracy and inherent biases, with teachers questioning its value in the learning process. This hesitancy from teachers suggests the need to build confidence in embracing generative AI and allay the anxieties about its impact on their work.

Yet, in the focus group data, students from refugee and migrant backgrounds seem to exhibit a proactive stance towards the utilisation of generative AI in their education, often learning about and initiating its use independently. As one recounts, “I was the one that started using it in class” (SFG). However, these students were aware of their knowledge gaps and expressed a strong desire for more information. Appreciating the potential for AI to positively impact their lives, they conveyed a need for assistance on effective use to be able to implement it: “We just need to know how to use it in a proper way” (SFG).

Interest in formalising learning about generative AI is apparent, with students indicating that tailored learning would increase their engagement and expertise: “The more you provide us with [an] AI course, the more interested we will be” (SFG). Even among those with some AI experience, there is an eagerness to further implement the technology, but this is held back by a knowledge gap.