CUHK–UCL international symposium

Gabriel Fan Fang

The Education University of Hong Kong

English-medium instruction (EMI) has been widely discussed and debated from various perspectives, particularly in relation to language policy and classroom practices. A key issue of EMI concerns the preparedness of and support for the main stakeholders, including both teachers and students. This talk reports on two studies conducted in the Chinese context and argues for the need to move beyond the four walls of the classroom to examine broader issues related to EMI. The findings suggest that many students are insufficiently prepared to achieve the dual goals of content and language learning, and that EMI programmes may widen the gap between students with different levels of English proficiency, potentially leading to educational inequality if not well-planned. Students also raised concerns about the quality of EMI teaching and highlighted the need for discipline-specific language support. In the era of artificial intelligence, this talk argues for a need of rethinking EMI through a decolonial lens in terms of how EMI and AI-mediated practices may be designed in order not to reinforce existing inequalities in language education.

Benjamin Luke Moorhouse

City University of Hong Kong

This talk frames EMI in the GenAI age through AI-mediated scholarship: the growing reliance on Large Language Model-powered “summarisers” and “chat-with-the-article” interfaces that insert algorithmic interpretation between readers and research. EMI students face many challenges when engaging with scholarship, including linguistic load, unfamiliar genres, dense argumentation, and inequitable access to disciplinary discourse, with GenAI often seen as a ‘solution’ to these difficulties.

Rather than positioning AI as the solution to these difficulties, I argue that we need be cautious of AI-mediated scholarship, and should instead facilitate and promote practices that scaffold authentic engagement (e.g., guided slow reading, annotation, dialogic discussion, and genre-aware reading routines) and that advocate for intellectual struggle as a condition for developing language skills, disciplinary knowledge and judgement. In doing so, we resist efficiency narratives and reclaims deep, critical reading as central to learning.

Christy Xuyan Qiu

The Hong Kong Polytechnic University

This presentation reports on a study investigating how task complexity, pragmatic demand, and interaction modality influence the use of discourse markers by second language (L2) learners from an English Medium Instruction (EMI) university in Hong Kong. The study investigated learners’ discourse marker use in interactions with both human interlocutors and ChatGPT across four distinct oral tasks. Frequencies of discourse marker use were categorised and analysed following Fung and Carter’s (2007) framework. Findings demonstrated that lower task complexity had a greater impact on the use of referential markers (which clarify or specify information) than on structural markers (which organise or sequence discourse). AI-mediated dialogues elicited increased use of sequential markers, and the combined effect of these factors shaped learners’ choices of referential and consequence markers. These results suggest that both task design and interaction modality influence EMI students’ discourse marker use, highlighting opportunities for teachers to leverage AI-mediated activities and task variation to promote more effective spoken communication in L2 contexts.

King Tat Daniel Fung

The University of Hong Kong

Language proficiency has always been an important variable in second language research and education. In English Medium Instruction (EMI), some researchers and practitioners are concerned about the English proficiency that students need in order to learn without experiencing too many difficulties. However, it is not always easy to gauge students’ academic English proficiency especially at the time when they transition from secondary schools to universities, because many students would not have taken an academic proficiency test such as IELTS. It is, therefore, of both theoretical and pedagogical value to develop an academic English proficiency assessment that is efficient and economical. This paper presents the development of such a test – an academic c-test – with the support of GenAI. The presentation will start with an introduction of the validity of c-tests, grounded in a systematic review, followed by a description of the development and evaluation of an academic c-test. Finally, research and pedagogical implications will be discussed.

Nathan Thomas

University College London

There are several key negotiations that often take place in EMI research: policy vs practice, outcomes vs experience, and, recently, tradition vs modernity in the way GenAI is incorporated, or not, in research and practice. We can also witness similar negotiations in current discussions of open science (a “modern” phenomenon) in relation to qualitative research, which has rich traditions. Fundamentally, is “modern” always preferential in relation to tradition? At what point does something modern become tradition? And is there a risk that practices associated with modernity or tradition become dogma? In thinking through some of these issues, I propose a research practice that appears anti-modern as a possible solution for the very modern dilemma qualitative researchers face in relation to open science: how can we become more “open” in our interpretive research practices? Participant Narratives are analytic artefacts that invite readers to engage directly with interpreted material, provide a fuller context than isolated excerpts, and offer a feasible alternative to accessible transcripts. Importantly, these narratives steer toward methodological rigor and ethical responsibility, without yielding to “modern” encroachment on the ethos of qualitative research. I demonstrate how they can be used by researchers interested in narrating the lived experience of EMI.

Sihan Zhou

The Chinese University of Hong Kong

The rapid expansion of generative AI (GenAI) has brought learner agency to the forefront of education research and practice. Given the transactional and reciprocal nature of human-AI interactions, learners must exercise agency not only through intentional efforts to direct interactions but also through simultaneous adaptation in response to AI-generated outputs. In this talk, I will introduce a recent project that investigates students’ agentic engagement with GenAI tools in English-medium instruction (EMI) contexts. In the first part of the talk, I will present a collaborative study that developed and validated an AI Agentic Engagement Scale (AAES). I will then share preliminary findings from a mixed-methods study that employed the AAES to examine EMI students’ agentic engagement and AI literacy in university contexts in China. Both studies highlight that agentic engagement with GenAI is grounded in critical, evaluative knowledge of the tools—rather than mere awareness of their functions. To conclude, I will discuss pedagogical implications for leveraging GenAI to promote agentic, strategic learning in EMI higher education settings.