Artificial intelligence (AI) is increasingly used in language education, yet little is known about how English as a foreign language (EFL) teachers’ emotions unfold over time as they integrate AI into their teaching. Guided by the circumplex model of affect and Bronfenbrenner’s bioecological model, this retrospective qualitative multiple-case study examined the emotional pathways of three Chinese university EFL teachers at different career stages. Data came primarily from semi-structured interviews, supplemented by retrospective reflective entries and platform-based teaching artefacts that clarified and contextualised reported incidents. Textual data were analysed through narrative reconstruction, coding at the episode level, interpretation of valence, arousal, and ecological conditions, and comparison across cases. The analysis identified three case-based interpretive pathways: volatility, characterised by high arousal and fluctuating valence; equilibrium, marked by moderate arousal and a stable positive orientation; and composure, reflected in low arousal and a neutral-to-positive orientation. These pathways were associated with case-specific configurations of teachers’ appraisals, interactions with students and colleagues, institutional arrangements, sociocultural narratives, career timing, and changing AI tool ecologies. The study contributes an ecological account of AI-related teacher emotion and highlights the value of context-sensitive, pathway-responsive support focused on technical reliability, pedagogical judgement, ethical use, and collegial learning.
Despite growing interest in note quality, understanding its relationship with lecture listening comprehension remains limited, particularly in multilingual and English-medium instruction (EMI) contexts. This study investigates how multilingual learners take notes during academic lectures and explores the relationship between note quality and lecture listening comprehension. Forty-one multilingual graduate students enrolled in a TESOL program in Canada participated in the study. Participants took notes using two methods, the structured (Cornell) and conventional, while listening to academic lectures. Data from a lecture listening comprehension test and participants’ notes were analyzed using Pearson’s correlation coefficient, descriptive statistics, and regression analysis, alongside content and semiotic analysis (Flick, 2014; Kress & van Leeuwen, 2006; Krippendorff, 2018). The findings show that multilingual learners generally demonstrated a solid understanding of the lecture content, despite some individual variation, and that note quality scores were positively correlated with lecture listening comprehension test scores. Participants drew on diverse linguistic resources in their notes, including their first language (L1), second language (L2) English, and translanguaging practices. They also employed a range of visual and spatial semiotic resources and note-taking strategies, including graphic organizers, arrows, underlining, headings, subheadings, circling, abbreviations, examples and sentence- and phrase-level notes. The findings underscore the importance of conceptualizing note-taking as a strategic, multilingual, and multimodal meaning-making process in EMI contexts. The study also points to the need for further research on how multilingual and semiotic note-taking practices relate to longer-term learning outcomes, as well as how explicit instructional support may influence note-taking effectiveness in EMI settings.
Productive teacher questioning is key to effective teaching and learning, and acquiring such questioning skills is a holy grail of teacher professional development. This study explored the contributions of observation-based mentoring to eight Hong Kong English language teachers’ evolving conceptualizations of effective classroom questioning by analyzing teacher-mentor interactions and their discursive representations. Data collected for the study comprised mentoring interactions in debriefing sessions following classroom observations and semi-structured interviews with the teachers. Analyses of the interactions revealed that the teachers became increasingly adept at using the conceptual terms and a heuristic framework of teacher questioning introduced in the mentoring program to represent and reason about their classroom questioning practices. Furthermore, the teacher-mentor interactions helped the teachers translate theoretical knowledge into questioning practices in the English classroom and led them to engage in self-initiated inquiries to hone their questioning skills. These findings have implications for language teachers and teacher educators regarding how mentoring can facilitate the development of classroom questioning strategies that enhance student engagement and foster a more learning-conducive environment.
Despite the growing popularity of task-based language teaching (TBLT) and assessment, the extent to which task-based tests promote TBLT preparation materials remains unexplored. To fill this gap, this study examines the alignment between East's (2021) four-criterion definition of task and materials designed for preparing learners to complete a task-based high-stakes exam in Hong Kong. Three sets of widely used test preparation English language textbooks were analysed in terms of task criteria. We found that less than thirty percent of the textbook activities fully met the definition of task across the series. However, forty percent satisfied three out of four task criteria, with communicative outcome being the least achieved criterion. In addition, more than half the sections provided language-related and exam-related information, indicating the existence of washback from the exam. We interpreted the overemphasis on examination strategies and information provision as negative washback. However, absence of explicit grammar instruction in the textbooks could be seen as a source of positive washback. Our findings highlight the challenge of introducing task-based materials, even where TBLT is the recommended pedagogic and testing approach.
The present study tracks the development of conversational implicature (CI) comprehension among adult Chinese L2 learners studying in the UK over an eight-month period and examines the extent to which learners’ motivation and L2 interaction predict gains in CI comprehension. Quantitative data were collected from 90 Chinese university students who completed two rounds of audiovisual CI comprehension tests and questionnaires administered before and after the eight-month interval. The questionnaires measured L2MSS, pragmatic motivation, willingness to communicate (WTC), social network characteristics, and L2 contact.The findings reveal significant gains in CI comprehension over time. Throughout the study-abroad period, participants maintained high levels of general motivation and pragmatic motivation, and overall WTC increased modestly over time. However, their interpersonal connections with English speakers weakened, and classroom-based English use eventually exceeded out-of-class interaction. Regression analyses indicate that pragmatic motivation was the only significant predictor of CI comprehension scores. Specifically, learners’ orientation toward accurate meaning conveyance and interpretation, together with their awareness of contextual information, plays a central role in the development of CI comprehension.
Generative AI (GenAI) has rapidly revolutionized educational practices. However, most existing studies do not integrate the evolution of CALL/MALL-to-GenAI chatbots into L2 writing instruction, and few review articles have examined this specific integration. To bridge these gaps, this scoping review synthesized emerging research in this area. It analyzed pedagogical innovations, implementation challenges, and future directions, drawing on metadata extracted from 101 studies (published 2012–2025) sourced from Web of Science (WOS), SpringerLink, ScienceDirect (Elsevier), Nature, EBSCO, JSTOR, and ACM. The main findings were that: (1) researchers in this area come primarily from China and the USA, with the remainder distributed across 30 countries and regions in Asia, North America, Europe, and Africa; (2) three research hotspots were identified, spanning from (Phase 1) cognitive-sociocultural mediation (2012–2019) to (Phase 2) the pre-GenAI automation era (2020–2022) and (Phase 3) the human-GenAI collaboration era (late 2022–2025), revealing a shift from mere tool usage toward a critical approach to integrating the human-AI relationship in L2 writing; (3) three key innovations were observed: metacognitive redesign of iterative feedback; human-AI symbiosis models that promote distributed cognition; and adaptive, stage-specific scaffolding across the writing process, collectively indicating a move from viewing GenAI as an assistive tool to regarding it as a generative partner; (4) nevertheless, persistent problems remain, including feedback inconsistency due to algorithmic hallucinations, an over-reliance on GenAI (compensatory writing) that may diminish critical thinking, and ongoing concerns regarding academic integrity and sociocultural bias. Research gaps, future directions, and recommendations for stakeholders are discussed.
Behavioural science offers systematic methods for identifying why educational practices occur and how they might be changed, yet these approaches remain underused in language-related education. This conceptual and methodological paper presents IDDEAS (Intervention Design, Delivery, Evaluation and Adoption System), an integrative framework that brings together the Behaviour Change Wheel, COM-B, the Theoretical Domains Framework (TDF), behaviour change techniques, APEASE criteria, and implementation and reporting tools. IDDEAS is operationalised through eight iterative stages: Identify, Define, Diagnose, Design, Deliver, Evaluate, Adapt and Adopt, and Share. The paper explains the logic of these stages and illustrates their relevance to digitally mediated English-medium instruction (EMI) in higher education. The EMI illustration draws on three linked studies from the Future of English project that used COM-B and/or the TDF to examine student engagement, teachers and support staff, and senior-management perspectives. These studies provide an empirical basis for the diagnostic stages of IDDEAS; the later stages are used to show how identified barriers and enablers can be translated into intervention design, delivery, evaluation, and institutional adoption. The framework offers applied linguistics researchers a structured way to connect behavioural diagnosis with pedagogical intervention and evaluation while retaining the contextual sensitivity required in multilingual educational settings.
Most research into instructional quality has been conducted in the context of STEM education, with issues relating to subject-specificity being addressed explicitly only recently. Subject-specific approaches might increase validity by capturing aspects of instructional quality that generic frameworks fail to acknowledge. This is particularly true for cognitive activation, which is largely dependent on the lesson content. We present psychometric evidence on the validity of a scale designed to assess students’ perceptions of cognitive activation in English as a foreign language education at primary level. Drawing on a sample of 513 German 4th graders (49% female, Mage = 10 years, SD = .59), we investigated the construct and criterion validity of a novel scale. We confirmed our theoretical model of communicative-cognitive activation, which included one general factor and six specific factors. Communicative-cognitive activation and its subdimensions are, to some extent, positively related to other generic dimensions of instructional quality, such as classroom management and student support. However, findings on the relation between communicative-cognitive activation and students’ EFL reading proficiency are inconclusive.
This special issue explores the opportunities, challenges, and ethical considerations arising from the integration of generative artificial intelligence (GenAI) into foreign/second language (L2) education. Drawing on fifteen articles published in this collection, we synthesize the field's evolving landscape across three interconnected dimensions: opportunities, challenges, and ethical considerations. We argue that GenAI holds great potential for scaffolding L2 learning and teaching processes, fostering learner engagement and positive emotions, supporting teacher development, and reshaping language assessment and educational innovation. At the same time, this collection foregrounds important tensions surrounding the reliability and contextual appropriateness of GenAI output, learner over-reliance, weakened agency, uneven access, and varying levels of teacher preparedness. Crucially, the ethical use of GenAI in L2 education demands attention not only to individual responsibility, but also to teacher mediation, institutional policy, assessment governance, data privacy, and educational equity. This editorial concludes by charting a research agenda that advocates theoretically grounded, empirically rigorous, and contextually sensitive inquiry into GenAI's role in transforming L2 education.
The emergence of artificial intelligence (AI) has brought innovative changes to language instruction by offering personalized and flexible educational opportunities. Although AI is widely recognized for delivering personalized feedback, its potential as a systematic, student-driven, mediated scaffolding tool that calibrates support to students' evolving writing needs remains underexplored. We therefore designed a student-driven AI-mediated scaffolding model (AI-MSM) for writing instruction and examined whether it was associated with gains in English as a foreign language (EFL) students' writing performance, engagement, and motivation. Forty Chinese EFL students were assigned to the experimental group and participated in a 10-session AI-MSM intervention, wherein they received implicit and explicit AI-mediated scaffolding to support their writing processes. Another 40 students received traditional writing instruction and served as the control group. Empirical data were collected through pre- and post-writing tests, engagement scales, motivation scales, and semi-structured interviews. The results of MANCOVA demonstrated that students in the AI-MSM condition showed greater gains in writing performance and engagement than those in the control group and exhibited preliminary trends in certain motivational dimensions under the present instructional arrangement. These findings suggest that AI-MSM may hold potential for addressing persistent challenges, such as insufficient personalized feedback in EFL writing instruction, although its implications for motivation warrant further investigation. To further explore students' perspectives on AI-mediated scaffolding, we conducted semi-structured interviews, which revealed students' positive perceptions of AI's potential as a mediated scaffolding tool in EFL writing instruction. The study provides a theoretical framework and practical guidelines for optimizing AI-assisted writing instruction.
Research on second language (L2) lexical inferencing has primarily examined cognitive and contextual factors, with limited attention to socio-emotional individual differences. Dynamic assessment research further emphasizes the value of examining not only static lexical inferencing performance but also learners’ inferencing potential under mediation. This study examines how empathy and executive function relate to initial lexical inferencing performance, lexical inferencing potential, and vocabulary retention from emotionally charged contexts. One hundred and two intermediate EFL undergraduates completed the Interpersonal Reactivity Index (IRI), a color-word Stroop task, a backward digit span task, and a computerized dynamic inferencing task featuring emotionally charged contexts, followed by a vocabulary retention test two weeks later. Multiple regression analyses examined the main effects, while exploratory correlations and hierarchical regression focused on IRI subdimensions. Results showed no significant associations between other-oriented empathy, a composite of Perspective Taking and Empathic Concern, and executive-function indicators in the main analyses. Exploratory analyses revealed that Empathic Concern was positively associated with longer reaction times on correct incongruent Stroop trials, while Personal Distress showed a marginal positive association. Other-oriented empathy positively predicted initial lexical inferencing performance. Follow-up analyses suggested that this association mainly reflected the shared contribution of Perspective Taking and Empathic Concern. In contrast, faster response times on correct incongruent Stroop trials significantly predicted lexical inferencing potential under mediation, whereas neither empathy nor executive-function indicators significantly predicted vocabulary retention. These findings highlight differential associations of socio-emotional dispositions and cognitive efficiency with initial lexical inferencing performance and modifiability under mediation.
In AI-enhanced language learning environments, understanding the drivers of sustained engagement remains a critical challenge. While previous research has identified various factors influencing technology acceptance, their complex interplay in shaping continuance intention requires further investigation. This study examines how teacher support, peer support, AI learning self-efficacy, and perceived enjoyment collectively influence learners’ continuance intention for AI-enhanced EFL learning. By integrating structural equation modeling (SEM) and fuzzy-set qualitative comparative analysis (fsQCA), the study moves beyond net effects to uncover conjunctural causality. Analyzing data from 705 Chinese university students, SEM results show that teacher support operates only indirectly via self-efficacy and enjoyment, whereas peer support exerts both direct and indirect effects. Complementing the SEM results, fsQCA reveals five equifinal configurations for high continuance intention, with perceived enjoyment as a core condition in most pathways. Notably, teacher and peer support can compensate for each other in specific contexts, a finding that linear models cannot capture. These insights advance technology acceptance theory by demonstrating that social, cognitive, and emotional factors combine in non-linear, configurational ways to sustain AI-enhanced language learning. Practical implications for designing adaptive support systems are discussed.
