명정희 Jeonghee Myeong , 박정아 Jeong-ah Park

DOI:10.15702/mall.2026.29.2.1 Vol.29(No.2) 1-29, 2026

This study investigates the capabilities and constraints of generative artificial intelligence in detecting and annotating Korean tense errors by comparing AI-generated error annotations with those produced by humans. Specifically, 589 tenseerror tokens from advanced Korean learners were annotated using ChatGPT-5.2. The analysis examined not only the concordance between AI and human annotations but also the accuracy of AI-generated corrections, pinpointing of error locations, and identification of error types. The findings revealed that 55.9% of the AI annotations were considered accurate, highlighting the potential of generative AI for annotating Korean tense errors. However, the AI often produced overcorrections, localized errors according to the corrected forms rather than the original error sites, and demonstrated inconsistent performance in distinguishing among omission, addition, and substitution errors. These results suggest that AIbased annotation requires more precise annotation guidelines, richer discourse context, and systematic attention to grammatical and semantic constraints on Korean tense forms. The study advances understanding of both the usefulness and boundaries of generative AI in annotating Korean learner errors by providing detailed analyses of correction accuracy, error location identification, and error type categorization.

generative artificial intelligence, learner error corpus, tense education, ChatGPT, error annotation

이혜진 Hyejin Lee

DOI:10.15702/mall.2026.29.2.30 Vol.29(No.2) 30-50, 2026

This study develops a Project-Based Service-Learning (PBSL) model that integrates vibe coding, a prompt-based generative AI development approach, into English teacher education coursework, positioning international student support as the authentic problem-solving context for pre-service English teachers. The PBSL model was implemented across two English education major courses at a non-metropolitan South Korean university. Over a semester, 39 participants designed and developed functional multilingual apps, brochures, and instructional videos addressing international students’ real campus needs, including dormitory navigation, cafeteria usage, academic email etiquette, counseling center access, and cultural adaptation, using vibe coding platforms. Each participant then conducted face-to-face English-medium mentoring sessions with international students to deliver outputs and collect user feedback. Drawing on two primary data sources, a pre-post survey questionnaires and open-ended reflective responses, the study examined competency changes and career identity formation among the participants. Paired-samples t-tests revealed statistically significant improvements across all five competency domains (all p < .001): creativity (d = 1.03), convergence (d = 0.94), global competence (d = 0.83), selfmanagement (d = 0.80), and self-efficacy (d = 0.77). Post-test standard deviations decreased consistently across all domains, suggesting that competency gains were observed across participants following the PBSL experience, rather than being concentrated among high-performing students. Thematic analysis of qualitative data further revealed that PBSL reshaped participant career identities across three dimensions: expanding the perceived value of English education from classroom instruction to multicultural mediation, enabling self-verification of career competencies through authentic user interaction, and broadening career pathways beyond the teacher-nonteacher dichotomy toward EdTech and related fields. The study validates vibe coding as an accessible, low-cost development tool enabling non-computer science majors to translate educational intentions into functional digital artifacts, and provides a replicable PBSL curricular framework for teacher education programs navigating declining enrollment, campus internationalization, and AI-driven digital transformation.

project-based learning, English education majors, service learning, generative AI, university internationalization

Sangmin-michelle Lee , Junseong Bang

DOI:10.15702/mall.2026.29.2.51 Vol.29(No.2) 51-68, 2026

This conceptual paper examines how Korean English as a Foreign Language (EFL) learners reconfigure their language use and attitudes through sustained engagement with generative artificial intelligence (GenAI). Using assemblage theory, the rewilding framework, and Thorne’s (2016, 2024) concept of morphologies of action as a foundation, the paper reimagines the interaction between learners and GenAI as a heterogeneous coupling. In this coupling, human and nonhuman actors collaborate to create new patterns of action. Synthesizing prior work on text-based autonomous use, the paper proposes four interlocking discourse features for empirical scrutiny: absence of repair sequences, scarcity of non-lexical markers, a recurrent three-step pattern, and turn-length asymmetry. These features constitute domesticated interaction. The paper extends the framework of Thorne and his colleagues by integrating the structural particularity of the Korean EFL ecology. It also theorizes a use-to-attitude pathway via a register convergence hypothesis. The paper’s central argument is that cumulative domesticated interaction, mediated by register narrowing, can reshape learners’ tacit models of communication, images of legitimate English users, and self-positioning. Distinguishing short-term gains from long-term sedimentation, the paper derives four rewilded design principles for the Korean EFL context.

Generative AI, assemblage theory, rewilding, morphologies of action, Korean EFL, domesticated interaction

이원지 Won Ji Lee , 이동주 Dong Ju Lee

DOI:10.15702/mall.2026.29.2.69 Vol.29(No.2) 69-90, 2026

This study designs, develops, and validates an AI-based Data-Driven Learning (DDL) web tool for elementary English learners and teachers. While DDL has shown strong pedagogical benefits through corpus-based inductive learning, its use in elementary contexts remains limited due to learners’ low proficiency, cognitive constraints, limited corpus access, and the complexity of existing tools. Meanwhile, artificial intelligence (AI) offers new opportunities but raises concerns regarding reliability, control, and ethics. To address these issues, this study proposes a controlled AI-based DDL tool (AIDDL) that maintains core DDL principles while ensuring learner appropriateness and instructional control. Using a designbased research approach, iterative cycles of design, implementation, analysis, and revision were conducted. AI-DDL was developed based on four principles: exploratory learning support, curriculum alignment, controlled AI use, and learnerfriendly design. It integrates textbook-based corpus data with selectively generated AI content. Key features include concordance exploration, collocation mapping, inference prompts, and storage and review functions. Evaluation by 11 experts showed high design validity (M = 4.69) and usability (M = 4.56). Revisions improved interface clarity, level adjustment, and system stability. The findings imply that AI-DDL is pedagogically sound and practically applicable, offering a viable model for controlled AI integration in elementary language learning.

Data-driven learning, AI-based web tool, elementary English education, corpus use, design-based research