A tool for teaching argumentative writing to EFL students

Teaching argumentative writing to tertiary-level learners of English as a Foreign Language (EFL) is a persistent challenge. Students often struggle to construct coherent, evidence-based arguments in their academic writing.

Their difficulties include a limited command of disciplinary vocabulary, over-reliance on simple sentence structures and a fragile understanding of how to structure claims, grounds and rebuttals convincingly. Traditional English for Academic Purposes (EAP) courses frequently separate language instruction from rhetorical training, leaving learners without a clear idea of how expert writers actually deploy grammar and lexis to achieve persuasive effect. 

• Spotlight guide: Bringing GenAI into the university classroom
• How I use AI tools to enhance EFL students’ speaking skills
• Three ways to develop students’ AI literacy

This gap is particularly noticeable for non-English majors, whose degree programmes rarely offer sustained, genre-focused writing support. To address this issue, we need an approach that makes authentic language patterns visible and manageable, while also guiding students toward critical, independent use of those patterns. This is not just important for assessment scores – the ability to construct sound arguments in English is a gateway skill for participation in global academic and professional discourse.

What did we do?

Our team developed an AI-powered corpus platform comprising over 900,000 words, drawn from expert opinion articles and argumentative essays, designed for use in argumentative writing training courses. The course was offered as a voluntary, non-credit extracurricular workshop, open to students from any discipline, and was not embedded in any existing degree curriculum. 

Twenty-four non-English-major undergraduates from diverse disciplines took a six-week course using the platform. The pedagogical design followed the Corpus-Based Language Pedagogy cycle, as outlined by Education University of Hong Kong’s Qing Ma, Jinglan Tan and Shanru Lin in their 2022 paper “The development of corpus-based language pedagogy for TESOL teachers”. Each week, teachers first introduced one component of philosopher S.E. Toulmin’s argumentative writing model, such as evidence, warrant or rebuttal, and then asked learners to explore the platform’s three core functions. 

The Keyword function allows students to look up how a word is actually used in real texts, helping them understand its meanings and typical contexts. The N-gram tool presents common word combinations tied to specific argument structures, ranked by frequency. The Top-K feature provides two types of suggestions in response to student queries, which are generated solely by GenAI or based on the corpus data respectively.

Learners were required to compare the two types of output, summarise patterns they observed in the corpus data and then apply those patterns in their own writing tasks. Importantly, three of the five writing tasks allowed platform use, but the first and last were completed without access to any tools, which ensured that any measured improvement reflected their internalised skill rather than a temporary boost due to the platform.

What outcomes did we see?

By the end of the training, students’ average essay marks had improved by nearly four points on a 30-point scale, demonstrating meaningful improvement in overall writing quality. The most significant gains were observed in language use and grammatical accuracy. Specifically, students who had previously relied on simple stance markers, such as “I think”, began to incorporate formal and genre-appropriate frames, like “it is widely argued that”. 

Furthermore, students’ final essays, written without any platform support, were clearly better than the earlier drafts produced with the tool’s help. Beyond linguistic enhancement, the potential of our training on critical thinking also emerged from the constant comparison between AI-generated and corpus-based responses. 

As learners were trained to treat GenAI as a brainstorming partner while using the corpus as a reliability check, they cultivated a sceptical attitude toward machine-generated texts. One participant observed that GenAI could spark ideas, but the corpus data often provided authoritative evidence. This practice transformed the platform from an answer-providing device into an environment for evaluative reasoning, counteracting the risk of passive reliance on GenAI tools. 

What can educators take away from our experience? 

Here are three practical steps for educators interested in trying a similar approach, with no custom platform required.

First, turn critical comparison into a routine. Providing both GenAI-generated responses and real-world texts gives students a reliable standard against which to judge language quality. Through comparisons, students are better equipped to spot potential inaccuracies or unnatural phrasing in the GenAI output, a simple habit that goes a long way in tackling AI hallucination.

Second, keep the role of teacher firmly in the loop. Let the technology serve as a resource, not a replacement for human instruction. 

Third, build evaluation into the task itself. For example, ask students to submit a short note with every draft: What did they take from GenAI? What did they change and why? This small step makes critical reflection part of the writing process, not an afterthought.

As the use of GenAI becomes more widespread, the pedagogical imperative will not be to ban or embrace it wholesale, but to frame its use within activities that reward searching, evaluating and refining rather than simply accepting.

By turning writing instruction into a cycle of guided discovery, we hope to move learners from passive dependence toward confident, evidence-based expression, a shift that holds value far beyond any single course.

Liu Kanlong is associate professor at Hong Kong Polytechnic University.

If you would like advice and insight from academics and university staff delivered direct to your inbox each week, sign up for the Campus newsletter.