For images, download the pdf.
How do I stop my students from writing with generative artificial intelligence (AI) in a way that does not reflect their thinking and their voices? This is a question we have heard from many educators in workshops that we have shared on the use of AI in writing. Often this larger question comes wrapped in concerns about plagiarism. This new technology has highlighted the urgent need to reimagine academic integrity and assessment practices (Coffey, 2024; Furze, 2024; Higgs & Stornaiuolo, 2024; Mcknight & Shipp, 2024; Merod, 2024; Payne et al., 2024; Trevithick, 2024). While we can imagine a future where AI detection software offers us the answer to all our problems, this is not our current reality. Most available technological ‘fixes’ are not yet equipped to accurately detect AI in written work (Coffey, 2024; Elkhatat et al., 2023). How then can English language educators move forward now? How might they support multilingual learners to resist the allure of a technology which promises better results without the work? Multilingual writers have insights to share that can disrupt harmful assumptions about student cheating and support the development of voice in writing.
Why does voice matter?
Canada’s educational system does little more than pay surface-level homage to cultural diversity and instead asks multicultural and multilingual people to assimilate (Olding, 2017). Whether the multilingual writers in our classrooms are children or adults, newcomers to Canada or naturalized Canadian citizens, are always considered ‘other’ (Olding 2017; Tajrobehkar, 2023). Language is one of the sign-posts used to indicate who can belong and who cannot (Skutnabb-Kangas, 1990; Olding 2017; Tajrobehkar, 2023). For those who can learn a language with ‘native-like’ proficiency, it is a source of social capital (Tajrobehkar, 2023). Multilingual educators in this study note the challenges both personally and for their students in achieving the authentic naturalness expected of them. One instructor stated:
I’m not judging some of my coworkers or something, but if we’re not all native speakers, sometimes it’s hard for even us to do that because we’re so trapped in how we have been writing … So you can’t really provide the variety or you can’t guarantee that what you provide is actually authentic or like feels natural when it comes to a native speaker … [We have to] measure up to this expectation of sounding authentic and sounding natural. (Participant 22)
For multilingual writers, their voices are contested spaces, not valued as integral and central to writing processes. Many process-oriented approaches to writing want nothing more than learners’ ‘voice’ but for multilingual writers, they are first being asked to transform their voices into the voice acceptable to the dominant culture (Mahboob & Szenes, 2010; Skutnabb-Kangas, 1989). Generative AI, while promising to support multilingual writers, offers fixes to standardize their language and produce formulaic texts (Asad et al., 2024; Evmenova et al., 2024; Sasaki, 2023; Wang, 2024) robbed of learners’ unique ways of languaging (Payne et al., 2024). While it is tempting to consider generative AI as a neutral tool, it is important to connect the tool to the world it has been introduced into (Mcknight & Shipp, 2024). Multilingual writers’ voices are often more valued when they can conform (Mcknight & Shipp, 2024; Skutnabb-Kangas, 1989). With such an expectation, there is a real risk of generative AI only reinforcing rather than challenging this system.
In our recent study using an inductive, qualitative approach to data collection and analysis (Charmaz, 2017), we interviewed 39 language and literacy educators from a variety of secondary and postsecondary settings about their experiences, questions and dilemmas surrounding teaching writing post generative AI.
This paper shares findings from a subset of participants, 15 multilingual educators whose teaching contexts included high school, college, university, teacher education, adult community English language education, and tutoring at a writing center (see Figure 1). While not all these educators exclusively taught English language learners, all of them had multilingual learners in their classrooms who they were supporting. As multilingual learners and educators, they had been negotiating similar, English dominant educational systems for much of their lives. Their experiences challenge assumptions about why multilingual students often rely on tools like generative AI in their writing.
Figure 1: Contexts of Multilingual Educators
Findings
For multilingual educators, plagiarism is less an issue of laziness, and more related to the pressure learners are under to “hide” themselves (Participant 22, English language educator) in English-spaces. Fundamentally, English language learners are asked to make a choice between “between sounding better or sounding like me” (Wang, 2024, p. 15). Multilingual educators understand this choice on a deeply personal level, and as a result offer strategies and approaches for elevating learners’ voices rather than reinforcing the damaging paradigms which suggest there is one correct way to ‘do English’ (Mahboob & Szenes, 2010; Tajrobehkar, 2023). While there are many variables outside their control in addressing systems of inequality, these educators shared two principles which oriented their practice. First, they focused on building community around writing practices. Second, educators found ways to reimagine assessment to point learners towards the importance of their voices.
Fostering relational writing communities
Fostering relationship between educators, learners, and the broader world was a central principle for these multilingual educators when responding to the complexities of generative AI and student voice. “I know my students, I know how they write and I know what their writing style or the skills that they have or they haven’t had yet, either better or worse,” (Participant 13, high school educator). Figure 2 illustrates a variety of ways in which these educators support developing writers: through focusing on learners’ goals, modelling, providing feedback, and creating opportunities to practice critical thinking.
Figure 2: How educators fostered relational writing communities
Learners’ goals: Multilingual educators were attuned to the social, cultural, and language-oriented goals and contexts of their students. For some, this looked like sharing religious, linguistic and cultural knowledge to support their understanding of student writing. Other participants focused on teaching genre or connecting classroom assignments to the future goals of their learners. One educator chose to become a licensed paralegal which “opened a new window to my teaching” (Participant 25) to address the questions and concerns of his learners, many of whom were newcomers to Canada. These educators prioritized learning about what their specific group of learners needed and finding ways to connect course content to these desires.
Modelling: Educators also focused on providing modelling to learners to ensure they had resources to meet the gaps in their own knowledge and skills. For some educators this involved editing checklists, notetaking, or oral language and reading. These educators provided modelling because they perceived learner use of generative AI as a learning gap. They understood learners might not have the skills they need to be able to represent their ideas or are simply afraid of making mistakes. While educators saw potential for generative AI to provide good modelling for their learners, they wanted to ensure students had options beyond generative AI to support their writing.
Feedback: Feedback for these multilingual educators was an opportunity for them to spend time getting to know their learners’ voices. Feedback was more than just an opportunity to correct student work, it was an opportunity to connect. For some educators this meant refocusing their feedback on content and ideation, rather than penalizing grammatical mistakes. For others, conferencing or daily journals were used to provide those opportunities for direct feedback and connection. This not only supported a deeper familiarity with learners’ voices but opened playful opportunities to discover misuse of AI during the process. For example, one participant described a playful approach to addressing student overuse of AI because he knew their voices well through his approach to providing feedback through writing conferences.
So they sometimes would come with these texts or just pieces of writing. And I notice that the vocabulary is a little bit sophisticated. It’s too serious, philosophical. I was like, Oh, when did you turn into a philosopher? And I start joking with them…My assessment is very often based on conferencing. (Participant 8)
Critical thinking: Finally, these educators understood writing as a process deeply related to thinking and found ways to facilitate and deepen learners’ connections to their writing through collaboration around shared texts, reading and oral language. For example, Participant 23 used texts about homelessness in Canada to engage her class in a broader discussion about their perceptions of homelessness and to challenge learners to engage with each other across differences. For Participant 16, collaboration looked like holding mock interviews where learners drilled interview questions with each other to prepare for interviews in the real world. Participant 33, a college instructor, discussed the importance of collaborative learning for de-centering the device and giving space for both strategies rather than relying on one way of acquiring information. What both collaboration and a focus on oral language and reading have in common is the way they facilitated a deeper connection between learners and between learners and educators. Through getting their learners thinking in and around their writing processes, they found learners were both developing as thinkers and writers.
Building in processes to support learners’ voices allowed educators to know who their learners were and what they needed for their futures. These educators understood that any time spent getting to know the students in their classrooms was time well-spent. “I feel like with the students, that’s the way that they learn best, is like having that conversation in a relationship piece and that part is what helps them to get better” (Participant 13, second language high school educator).
Reimagining assessment
Cultivating relationship with learners wasn’t the only way educators centered the voices of multilingual writers. It was also necessary to align their assignments and assessment methods to reflect these values. Redesigning their assignments and assessments involved incorporating process-oriented approaches to writing instruction in ways that often included multimodal text making.
|
Multimodal Assignments and Assessment (P34, P32, P8, P3, P2) |
Process-Oriented Approaches (P34, P33, P32, P29, P18, P16, P22, P8, P3) |
|
Visualization (P8) |
Focus on critical thinking in assessment criteria and assignment design (P34, P33, P32, P29, P18, P8, P3) |
|
Graphic organizers (P2, P3) |
Focus on structure (P34, P32, P18, P8, P33, P29, P16, P22, P23) |
|
Connecting writing to images (P32, P34) |
Building writing in stages (P8, P34) |
|
Movement (P8) |
Modelling their own thinking/writing process (P29) |
Figure 3: Multimodal Assignments and Process-oriented Approaches
Multimodal assignments and assessment
A number of these educators talked about the need to shift assessments and assignments to include multimodal elements rather than focusing exclusively on writing (see Figure 3). Participant 8 used visualization as a brainstorming tool to support his high school learners: “So the visualization allows for that type of enriching of ideas before we … actually translate it into writing. And then the writing has to happen also in a gradual way”. Graphic organizers were also helpful in supporting developing writers and helped learners to “get it [the writing process] better”. Two of the multilingual educators talked about assignments which required learners to generate or connect to images. For example, participant 32 had learners create a social media campaign and create images with generative AI when they could not find the appropriate images without it. Participant 34 centered presentations instead of written assignments to increase engagement. “Even if they’re not doing that [engaging] through writing, they have other ways that they can engage”. Movement was also something participant 8 used to “activate the neurons”. These examples point to the connections between writing and other types of texts. Research suggests multimodal texts allow opportunities to consider and reconsider the text to increase engagement (Kang & Yi, 2023; Liu et al., 2024). However, this translation between and across modes is best suited for slow and process-oriented approaches which allow time for revision and reconsideration (Jacob et al., 2023; Liu et al., 2024).
Process-oriented approaches
Process approaches to teaching writing were crucial to supporting the development of learners’ voices. For these educators it meant focusing on critical thinking, teaching structure, building writing in stages, and modelling their own thinking and writing processes. To focus on critical thinking meant shifting assessment criteria to highlight the elements of an assignment focused on argumentation, reflection or problem-solving. Some educators focused on teaching genre and audience as a way to focus on structure. Critically, these educators found ways to slow down and build up to assignments. Participant 34 notes that post-generative AI “my teaching is a little bit more detailed because I want to make sure that they get it and I want to see a sample of what they can produce before they write the assignment.” For another educator, writing is a gradual process: Teachers will tell students to write a short essay, a three-page essay on the following topic.
That is a big mountain for them. So what I do is, for example, we start with the let’s do an Instagram post. So they know that is very short. So they kind of draw it and so on in a small box. So they have a small paragraph there. And then we will expand that into how about we do a small paragraph and then a longer paragraph and we expand, expand until we get to the 1.5 pages that I need at the end. So it goes, it has to be gradual. (Participant 8)
Methods for breaking the writing process into manageable pieces were also explored by modelling the thinking process. In sharing their strategy for paragraph writing, participant 29 notes: “It’s kind of like walking my students through my thinking process. So, I sometimes write as I talk to them”. Through slowing down, educators point out all the elements which go into constructing a piece of writing and the time it takes to develop a clear writer’s voice. Despite the constraints on their time, these educators articulated the importance of slowing down and supporting writing through multimodal elements to ensure all learners are supported and engaged in the writing process.
What does this mean for academic integrity?
Relational classrooms, multimodal assignments and process-approaches to writing instruction and assessment help to solve the problem of plagiarism that has been accelerated by generative AI and to support the development of voice and critical thinking in multilingual students. While there is no quick fix, what these strategies point towards is the need to address the root of the problem, not just the branches. The multilingual educators in this study teach us that plagiarism is not simply about laziness or learners not wanting to do the work. For multilingual writers, it is a symptom of a deeper sense of inadequacy and a byproduct of an educational system which teaches them their voices do not matter.
Conclusion
Generative AI is here to stay. For multilingual writers who are already facing the pressure to write in correct ways, it is important to consider how generative AI builds confidence and agency instead of diminishing it (McKnight & Shipp, 2024; Payne et al., 2024; Smith, 2024). These multilingual educators challenge us to question why multilingual learners (mis)use generative AI and inspire us to facilitate relationally rich classrooms. The good news is the practices shared by these educators are not revolutionary. Many of these educators have been using these strategies to support their learners well before the introduction of generative AI. For those of us working in classrooms to support culturally diverse and multilingual writers, these educators remind us that while AI certainly introduces new complexities, slowing down and building relationship with our learners remains at the core of our work. With these principles grounding our hopes for our students, we may see them “bringing themselves into their writing” (Participant 13), playing with their knowledge of both cultures (Participant 8), and using generative AI to help elevate their voices, rather than hide them.
References
Asad, M. M., Shahzad, S., Shah, S. H. A., Sherwani, F., & Almusharraf, N. M. (2024). ChatGPT as artificial intelligence-based generative multimedia for English writing pedagogy: Challenges and opportunities from an educator’s perspective. The International Journal of Information and Learning Technology, 41(5), 490–506. https://doi.org/10.1108/IJILT-02-2024-0021
Charmaz, K. (2017). The power of constructivist grounded theory for critical inquiry. Qualitative Inquiry, 23(1), 34–45. https://doi.org/10.1177%2F1077800416657105
Coffey, L. (2024). Professors cautious of tools to detect AI-generated writing. Inside Higher Education. https://www.insidehighered.com/news/tech-innovation/artificial-intelligence/2024/02/09/professors-proceed-caution-using-ai
Coffey, L. (2024). Students and professors believe AI will aid cheating. Inside Higher Ed. https://www.insidehighered.com/news/tech-innovation/artificial-intelligence/2024/07/29/students-and-professors-expect-more
Elkhatat, A. M., Elsaid, K., & Almeer, S. (2023). Evaluating the efficacy of AI content detection tools in differentiating between human and AI-generated text. International Journal for Educational Integrity, 19(1), 17–16. https://doi.org/10.1007/s40979-023-00140-5
Evmenova, A. S., Regan, K., Mergen, R., & Hrisseh, R. (2024). Improving writing feedback for struggling writers: Generative AI to the rescue? TechTrends, 68(4), 790–802. https://doi.org/10.1007/s11528-024-00965-y
Furze, L. (2024). Generative AI, plagiarism, and “cheating”. Leon Furze. https://leonfurze.com/2023/09/20/generative-ai-plagiarism-and-cheating/
Higgs, J. M., & Stornaiuolo, A. (2024). Being human in the age of generative AI: Young people’s ethical concerns about writing and living with machines. Reading Research Quarterly, 59(4), 632–650. https://doi.org/10.1002/rrq.552
Jacob, S., Tate, T., & Warschauer, M. (2023). Emergent AI-assisted discourse: Case study of a second language writer authoring with ChatGPT. Computation and Language, 1–24. https://doi.org/10.48550/arxiv.2310.10903
Kang, J., & Yi, Y. (2023). Beyond ChatGPT: Multimodal generative AI for L2 writers. Journal of Second Language Writing, 62, 1–2. https://doi.org/10.1016/j.jslw.2023.101070
Liu, M., Zhang, L. J., & Biebricher, C. (2024). Investigating students’ cognitive processes in generative AI-assisted digital multimodal composing and traditional writing. Computers and Education, 211, 1–2. https://doi.org/10.1016/j.compedu.2023.104977
Mahboob, A., & Szenes, E. (2010). Linguicism and racism in assessment practices in higher education. Linguistics and the Human Sciences, 3(3), 325–354. https://doi.org/10.1558/lhs.v3i3.325
McKnight, L., & Shipp, C. (2024). “Just a tool”? Troubling language and power in generative AI writing. English Teaching: Practice and Critique, 23(1), 23–35. https://doi.org/10.1108/ETPC-08-2023-0092
Olding, L. (2017). Racism and English language learning: Employing an anti-racist approach to english as an additional language education. SFU Educational Review, 9. https://doi.org/10.21810/sfuer.v9i.310
Payne, A. L., Austin, T., & Clemons, A. M. (2024). Beyond the front yard: The dehumanizing message of accent-altering technology. Applied Linguistics, 45(3), 553–560. https://doi.org/10.1093/applin/amae002
Sasaki, M. (2023). AI tools as affordances and contradictions for EFL writers: Emic perspectives and L1 use as a resource. Journal of Second Language Writing, 62, 101068-. https://doi.org/10.1016/j.jslw.2023.101068
Skutnabb-Kangas, T. (1989). Multilingualism and the education of minority children. Estudios Fronterizos, 7(8), 36–67. https://dialnet.unirioja.es/servlet/articulo?codigo=5196227
Skutnabb-Kangas, T. (1990). Legitimating or delegitimating new forms of racism – the role of researchers. Journal of Multilingual and Multicultural Development, 11(1–2). https://doi.org/10.1080/01434632.1990.9994402
Smith, J. M. (2024). “I’m Sorry, but I Can’t Assist”: Bias in Generative AI. Proceedings of the 2024 on RESPECT Annual Conference, 75–80. https://doi.org/10.1145/3653666.3656065
Tajrobehkar, B. (2023). Orientalism and linguicism: how language marks Iranian-Canadians as a Racial “other.” International Journal of Qualitative Studies in Education, 36(4), 655–671. https://doi.org/10.1080/09518398.2021.1885069
Trevithick, M. (2024). Western University students continue to cheat but few getting caught using AI to do it. CBC. https://www.cbc.ca/news/canada/london/western-university-students-continue-to-cheat-but-few-getting-caught-using-ai-to-do-it-1.7400216
Wang, C. (2024). Exploring students’ generative AI-assisted writing processes: Perceptions and experiences from native and nonnative English speakers. Technology, Knowledge and Learning. https://doi.org/10.1007/s10758-024-09744-3
Author Bios
Mercedes Veselka is a PhD student at York University. She is also an English language instructor who has primarily worked in community-based LINC programs. Her research explores educators’ experiences bringing theory into practice in adult literacy and language education.
Dr. Mary Ott is an assistant professor of literacy in the Faculty of Education at York University. Her research explores agency in multimodal literacy practices and pedagogies.