Exploring the Digital Transformation of Generative AI-Assisted Foreign Language Education: A Socio-Technical Systems Perspective Based on Mixed-Methods

Abstract

This study investigates the complex dynamics and impacts of generative AI integration in foreign language education through the lens of the Generative AI-assisted Foreign Language Education Socio-Technical System (GAIFL-STS) model. Employing an integrated mixed-methods design, the study combines qualitative case studies and hybrid simulation modeling to examine the affordances, challenges, and implications of AI adoption from a multi-level, multi-dimensional, and multi-stakeholder perspective. The qualitative findings, based on interviews, observations, and document analyses, reveal the transformative potential of generative AI in enhancing language learning experiences, as well as the social, cultural, and ethical tensions that arise in the process. The quantitative results, derived from system dynamics and agent-based modeling, provide a systemic and dynamic understanding of the key variables, feedback loops, and emergent properties that shape the trajectories and outcomes of AI integration. The integrated findings offer valuable insights into the strategies, practices, and policies that can support the effective, equitable, and responsible implementation of AI in language education.

1. Introduction

The rapid advancements in artificial intelligence (AI) have profoundly reshaped various domains of human activity, including education. In particular, the emergence of generative AI (GenAI) techniques, such as large language models [1], knowledge distillation [2], and transfer learning [3], has opened up new possibilities for personalized, interactive, and immersive learning experiences. Foreign language education, a field that heavily relies on authentic communication and contextualized practice, stands to benefit significantly from these AI-powered innovations [4,5].

However, despite the growing interest and investment in AI-assisted language learning, the current research landscape remains fragmented and limited in scope. Most existing studies have focused on narrow applications of AI, such as chatbots [6], pronunciation feedback [7], or essay scoring [8], without considering the broader implications for educational processes, structures, and outcomes. Moreover, there is a lack of theoretical frameworks that can guide the systematic investigation and integration of AI in foreign language education, taking into account the complex interplay of technological, pedagogical, and contextual factors [9,10].

To address these gaps, this study proposes a novel conceptual model, the Generative AI-assisted Foreign Language Education Socio-Technical System (GAIFL-STS), to examine the digital transformation of foreign language education in the era of GenAI. Drawing upon the socio-technical systems theory [11,12] and the TPACK framework [13], the GAIFL-STS model posits that the successful integration of GenAI in foreign language education requires the dynamic alignment and reconfiguration of three interdependent subsystems: the technical subsystem (AI affordances and constraints), the organizational subsystem (institutional strategies, policies, and practices), and the human subsystem (teachers’ and learners’ attitudes, beliefs, and behaviors).

Through the lens of the GAIFL-STS model, this study aims to investigate three key research questions:

• (1). What are the affordances and challenges of GenAI technologies for foreign language teaching and learning, as perceived by teachers and learners?

• (2). How do educational institutions adapt their strategies, policies, and practices to support the integration of GenAI in foreign language programs?

• (3). To what extent and under what conditions can GenAI enhance the effectiveness, efficiency, and equity of foreign language education, and what are the unintended consequences and ethical implications?

To answer these questions, this study adopts an integrated mixed-methods design that combines an embedded single-case study with a hybrid simulation approach, involving the concurrent and iterative collection, analysis, and integration of qualitative and quantitative data [14]. The qualitative phase employs an embedded single-case study design, focusing on a representative Chinese university that has adopted generative AI tools in three foreign language courses. These courses serve as multiple units of analysis within the single case, whose data are analyzed using a thematic analysis and constant comparative methods to identify the affordances, challenges, and implications of AI integration, as well as the strategies and practices employed by educational institutions to support AI adoption [15]. The quantitative phase involves the development and application of a hybrid simulation model that combines system dynamics (SD) and agent-based modeling (ABM) to capture the complex dynamics and emergent properties of the AI-assisted language education system. The simulation results are then integrated with the qualitative findings to generate holistic and contextualized insights into the digital transformation process and its implications.

By iteratively integrating the quantitative and qualitative findings, this study seeks to provide a comprehensive, nuanced understanding of the digital transformation in generative AI-assisted foreign language education. The results can inform the development of evidence-based policies, practices, and technologies to support the effective, equitable, and ethical implementation of AI in language education. Moreover, the GAIFL-STS model and mixed-methods design can serve as a methodological framework for future research on AI-enhanced education across diverse domains and contexts.

The rest of the paper is structured as follows. Section 2 reviews the relevant literature on AI in education and foreign language learning, identifies research gaps, and proposes the GAIFL-STS model as a conceptual framework. Section 3 describes the mixed-methods research design and gives an analysis of the results of the study. Section 4 discusses the results of the study and the main theoretical contributions. Section 5 summarizes the main findings of the paper, acknowledges the limitations of the study, and gives practical recommendations guided by the paper’s conclusions.

2. Problem Description and Theoretical Framework

2.1. Defining the Digital Transformation of Generative AI-Assisted Foreign Language Education

The advent of GenAI, powered by advanced techniques such as large language models, knowledge distillation, and transfer learning, has ushered in a new era of digital transformation in foreign language education [1]. This transformative process entails the fundamental restructuring of pedagogical practices, organizational processes, and learner experiences through the strategic integration of AI-based tools and systems [16]. While extant research has explored the applications of AI in education, such as intelligent tutoring systems and adaptive learning environments [17], the specific implications of GenAI for foreign language teaching and learning remain understudied.

Drawing upon the socio-technical systems perspective [18], we conceptualize the digital transformation of generative AI-assisted foreign language education as a multi-dimensional, evolutionary process that involves the alignment and reconfiguration of technological affordances, organizational structures, and human agencies. On the technical dimension, GenAI enables the creation of personalized learning content, real-time feedback, and immersive language practice through natural language processing, dialog generation, and speech synthesis [19]. These AI-powered affordances have the potential to redefine the roles and competencies of foreign language teachers, shifting their focus from content delivery to learning design and facilitation [20].

On the organizational dimension, the implementation of GenAI in foreign language education requires the adaptation of institutional strategies, policies, and infrastructure to support the seamless integration and ethical governance of AI technologies [21]. This may involve the redesign of curriculum, assessment, and professional development programs to align with the AI-enhanced pedagogical approaches [22]. Moreover, the adoption of GenAI may challenge the existing power dynamics and communication patterns among educational stakeholders, necessitating the cultivation of a collaborative, innovative organizational culture [23].

On the human dimension, the digital transformation of generative AI-assisted foreign language education is ultimately realized through the agentic actions and interactions of teachers and learners in situated contexts. Teachers’ technology acceptance, pedagogical beliefs, and AI literacy significantly shape their willingness and ability to leverage GenAI for instructional purposes [24]. Learners’ self-regulated learning strategies, metacognitive skills, and trust in AI systems also play a crucial role in determining the effectiveness and sustainability of AI-assisted language learning [25].

Importantly, the digital transformation of generative AI-assisted foreign language education is not a linear, deterministic process but an emergent, iterative one that unfolds through the dynamic interplay of technological, organizational, and human factors over time. As GenAI technologies continue to evolve and scale, their affordances for foreign language education may manifest in unexpected ways, creating new opportunities and challenges for innovation and inclusion [26]. Therefore, to fully capture the complexity and nuances of this transformative phenomenon, we propose the GAIFL-STS model as a conceptual framework to guide future research and practice.

2.2. GAIFL-STS Model: Generative AI-Assisted Foreign Language Education Socio-Technical System Model

Grounded in the socio-technical systems theory [27], we propose the GAIFL-STS model to capture the complex interplay among technology, organization, and human actors in the digital transformation of foreign language education. The GAIFL-STS model extends Lyytinen et al.’s (2021) socio-technical affordance theory [28], providing a multi-level, dynamic perspective to investigate the evolutionary process and impact mechanism of GenAI in reshaping foreign language teaching and learning.

As illustrated in Figure 1, the GAIFL-STS model consists of three interconnected subsystems: (1) the technical subsystem, which encompasses the GenAI techniques (e.g., large language models, knowledge distillation, transfer learning) and their affordances for personalized content creation, intelligent tutoring, and adaptive assessment; (2) the organizational subsystem, involving the structural, processual, and cultural changes in educational institutions to facilitate the integration and governance of AI technologies; and (3) the human subsystem, including foreign language teachers, learners, and administrators who interpret, appropriate, and enact the AI-enabled affordances in situated practices [29].

The GAIFL-STS model posits that the digital transformation of foreign language education unfolds through the recursive interactions among the three subsystems. On the one hand, the development and deployment of GenAI technologies in foreign language education are deeply shaped by the organizational contexts and human agencies. For instance, teachers’ technology beliefs, innovation orientations, and professional competencies significantly influence their adoption and adaptation of AI-based tools in pedagogical practices [30]. Learners’ self-regulated learning strategies, AI literacy, and trust in AI systems also play a crucial role in realizing the potential of GenAI for personalized learning [31].

On the other hand, the introduction of GenAI into foreign language education, in turn, triggers the restructuring of organizational processes, roles, and norms, as well as the reconfiguration of human cognition, affect, and behavior. For example, the increasing use of AI-powered chatbots and virtual agents for language practice may lead to the redefining of teacher roles from knowledge transmitters to learning facilitators and the re-engineering of teacher-student interaction patterns [32]. The AI-enabled multimodal feedback and learning analytics may foster learners’ metacognitive awareness, emotional engagement, and self-efficacy in language acquisition [33].

Moreover, the GAIFL-STS model highlights the dynamic, emergent nature of the digital transformation process. As GenAI technologies continue to advance and scale, their affordances for foreign language education may evolve in unexpected ways, creating new opportunities and challenges for organizational adaptation and human development. The socio-technical alignments among the three subsystems are not static or deterministic but are constantly negotiated and recalibrated through the reflective actions of multiple stakeholders [34].

Therefore, to fully unpack the transformative potential and pitfalls of GenAI in foreign language education, researchers need to adopt a holistic, contextualized approach that attends to the intricate interplay of technological features, organizational structures, and human experiences. The GAIFL-STS model provides a conceptual foundation for such an integrative inquiry, guiding the formulation of research questions, design of empirical studies, and interpretation of findings to advance our understanding of this emerging phenomenon.

2.3. Validity and Operationalization of Mixed Research Methods

To comprehensively investigate the complex, multi-faceted phenomenon of digital transformation in generative AI-assisted foreign language education, we propose a mixed-methods research design that integrates qualitative and computational quantitative approaches [35]. Mixed-methods research has gained increasing traction in educational research due to its potential to provide a more holistic, nuanced understanding of the interplay between technology, pedagogy, and learning [36].

The integration of qualitative research and simulation modeling offers unique opportunities to explore the dynamic, emergent nature of socio-technical interactions in AI-assisted language education [37]. Qualitative methods, such as in-depth interviews, focus groups, participant observations, and document analyses, provide rich, contextualized insights into the lived experiences, perceptions, and practices of various stakeholders engaging with GenAI tools [38]. These methods shed light on the affordances, challenges, and implications of AI adoption, as well as the social, cultural, and ethical factors that shape the integration of AI in language programs [39].

Simulation modeling, on the other hand, offers a systemic and dynamic perspective on the complex interactions and emergent properties of the AI-assisted language education system [40]. By combining system dynamics and agent-based modeling approaches, we can identify key variables, feedback loops, and tipping points that contribute to the trajectories and outcomes of AI integration, as well as explore long-term patterns and potential consequences under different scenarios and policies [41].

While simulation modeling is not empirical research per se, it is grounded in empirical data and theoretical foundations [42]. In our study, the qualitative findings inform the development, parameterization, and calibration of the simulation model, ensuring that the model is anchored in the real-world experiences and practices of educational stakeholders [43]. The simulation model, in turn, provides a virtual laboratory to test and refine the insights derived from the qualitative analysis, enhancing their internal consistency and logical coherence [44].

The epistemological underpinning of this mixed-methods approach lies in the pragmatic paradigm, which emphasizes the problem-centered, pluralistic, and real-world-oriented nature of research [45]. Pragmatism acknowledges the value of both qualitative and simulation methods in generating actionable knowledge and informing decision-making processes [46]. By combining these methods, we can leverage their complementary strengths to develop a more holistic, nuanced, and practically relevant understanding of the complex dynamics and impacts of GenAI integration in foreign language education [47].

To operationalize the mixed-methods research design, we adopt an integrated strategy that involves the concurrent and iterative collection, analysis, and integration of qualitative and quantitative data [48]. In the qualitative phase, in-depth interviews, participant observations, and document analyses are conducted to explore the perceptions, experiences, and practices of key stakeholders in three representative foreign language courses that have adopted GenAI tools. The qualitative data are analyzed using a thematic analysis and constant comparative methods to identify the key affordances, challenges, and implications of AI integration, as well as the strategies and practices employed by educational institutions to support AI adoption [15].

In the quantitative phase, a hybrid simulation model combining system dynamics and agent-based modeling is developed to capture the complex dynamics and emergent properties of the AI-assisted language education system. The model is parameterized and calibrated using empirical data from the qualitative findings, as well as relevant literature and expert opinions. Sensitivity analyses and scenario testing are conducted to examine the impact of key variables and interventions on the system behavior and learning outcomes over time. The simulation results are then integrated with the qualitative findings to generate holistic and contextualized insights into the digital transformation process and its implications.

By iteratively integrating the qualitative and quantitative findings, the mixed-methods approach provides a more comprehensive and nuanced understanding of the complex dynamics and impacts of GenAI integration in foreign language education. The GAIFL-STS model serves as a conceptual and methodological framework to guide the systematic investigation and integration of multi-level, multi-dimensional, and multi-stakeholder perspectives on AI-assisted language learning. The findings and insights generated from this study can inform the development of evidence-based policies, practices, and technologies to support the effective, equitable, and responsible implementation of AI in language education, while the mixed-methods research design can be adapted and extended to investigate AI-enhanced education in other domains and contexts.

3. Research Process and Results

3.1. The Qualitative Study: An In-Depth Description of the Whole Process of Using GenAI for Foreign Language Teaching and Learning

3.1.1. Qualitative Research Design

The present study adopts an embedded single-case study design, combining various data collection methods such as in-depth interviews, participant observation, and document analyses to comprehensively and deeply explore the application process and impact of GenAI in foreign language teaching. The choice of an embedded single-case study design is based on its advantages in examining a contemporary phenomenon within its real-world context, especially when the boundaries between phenomenon and context may not be clearly evident [49]. As Yin (2009) argues, an embedded case study design allows for the investigation of a single case with multiple units of analysis, providing a more detailed and nuanced understanding of the case [50].

The selection of research objects adopts purposive sampling to ensure the diversity and representativeness of the sample. Adeoye-Olatunde (2021) emphasized that in qualitative research, purposive sampling can help researchers select information-rich cases to gain an in-depth understanding of the core of the research question [51]. Based on this principle, we selected three representative case courses from the School of Foreign Languages at Harbin Institute of Technology: English and American Literature, Comparative Literature and World Literature, and Western Literary Theory. These courses represent different types of courses in foreign language education, involving different focuses such as language skill cultivation, literary appreciation, and theoretical analysis. For each case course, we tracked and observed a complete semester (16 weeks) to capture the entire process of the GenAI application. This long-term immersive research design aligns with Lewis’s (2015) suggestion that qualitative research should involve long-term, intensive data collection in natural settings [52].

The research participants included 9 foreign language teachers (3 for each case course, including 1 main lecturer and 2 teaching assistants), 45 students (15 for each case course, covering different grades and academic performances), 3 educational administrators (including school leaders and academic affairs staff), and 3 AI developers (technical staff responsible for developing and maintaining the GenAI system). This diverse composition of participants enables us to understand the application of GenAI in foreign language teaching from different perspectives and levels, which is consistent with Hays and McKibben’s (2021) view that multiple case unit analysis studies should focus on “multivoicedness” [53].

Data collection employs a strategy that combines multiple methods, including in-depth interviews, participant observation, and document analyses. This methodological triangulation design aims to enhance the richness and credibility of the data [54]. Specifically, we conducted 1–2 semi-structured in-depth interviews with each participant, each lasting 60–90 min. The interview outline was designed based on the GAIFL-STS model, covering aspects such as technical characteristics, teaching practices, learning experiences, and organizational support. As Roberts (2020) suggested, we adopted the “three-step interview method,” focusing on life history, current experience details, and reflection on meaning [55]. In terms of participant observation, researchers participated in classroom teaching 1–2 times per week, observing and recording the specific application methods of GenAI and teacher-student interactions. This “insider” perspective can help us capture details and implicit meanings that may be overlooked in formal interviews [56]. The document analysis includes collecting and analyzing documents such as course syllabi, teaching materials, student assignments, and AI system logs, providing us with important background information and supplementary evidence.

To ensure the ethicality and rigor of the research, we have taken several measures. First, informed consent was obtained from all participants through the class monitors, who served as the representatives of the students in each class. Due to the nature of the study, which mainly used a survey questionnaire and did not involve sensitive data, and the fact that the questionnaire was distributed to all classes taught by the author in English courses, obtaining individual consent from each participant was deemed impractical. The class monitors collected the consent of all participants and signed the informed consent form on their behalf, as some participants, including doctoral students, were not present on campus during the data collection period. The informed consent form clearly stated the purpose, procedures, and potential risks and benefits of the study, as well as the voluntary nature of participation and the measures taken to ensure confidentiality and anonymity. It is important to note that the ethical approval for this study was waived by the IRB of the Faculty of Humanities and Social Sciences at Harbin Institute of Technology. The waiver was granted based on the minimal risk nature of the study, which involved no more than survey procedures and did not collect sensitive or identifiable information from the participants.

Despite the waiver of ethical approval, we have taken additional measures to ensure the ethicality and rigor of the research. Anonymization was applied during the data collection process to protect participants’ privacy. All personal identifiers were removed from the questionnaires, and the data were stored securely in password-protected files accessible only to the research team. Furthermore, we adopted the “trustworthiness criteria” to enhance research quality, including strategies such as prolonged engagement, persistent observation, triangulation, peer debriefing, and member checking [57]. For example, we invited two qualitative research experts who were not involved in the study to review the analysis process and results, providing an external audit of the research procedures and findings. We also conducted member checking by presenting the preliminary analysis results to a subset of participants and obtaining their opinions and feedback, which helped validate the accuracy and credibility of the interpretations.

3.1.2. Research Process

In this study, we employed a thematic analysis to systematically examine the rich qualitative data collected through various sources, including classroom observations, in-depth interviews, document analyses, participatory observation, focus groups, and student reflective journals. A thematic analysis is a widely used method in qualitative research that allows for identifying, analyzing, and reporting patterns (themes) within data [58]. Its flexibility and theoretical adaptability make it suitable for both inductive and deductive approaches, enabling researchers to generate themes from the data or to analyze the data through the lens of existing theoretical frameworks [59].

We followed the six-phase thematic analysis process proposed by Braun and Clarke (2006), which has been widely recognized for its rigor and applicability in educational technology research [60]. The six phases are: (1) familiarizing with the data, (2) generating initial codes, (3) searching for themes, (4) reviewing themes, (5) defining and naming themes, and (6) producing the report. Throughout this process, we utilized NVivo 12 software to manage the large volume of qualitative data and to facilitate the coding and theme development process [61].

In the initial phase, we immersed ourselves in the data by repeatedly reading through the transcripts, observation notes, and documents to gain a comprehensive understanding of the data’s breadth and depth. We actively engaged in memo-writing to record our initial impressions and potential coding ideas [62].

Next, we proceeded to generate initial codes by systematically coding interesting features of the data and collating relevant data extracts under each code. We adopted an open coding approach, allowing codes to emerge inductively from the data without imposing a pre-existing coding framework [63]. To ensure consistency and comprehensiveness, three researchers independently coded the same subset of data and then compared and discussed their coding until a consensus was reached.

Table 1 presents a sample of the initial coding process, illustrating how data extracted from various sources were assigned codes that capture their essential meaning (See Appendix A, Table A1 for details).

After the initial coding, we shifted our focus to the broader level of themes, collating codes into potential themes and gathering all relevant coded data extracts within each theme. We employed thematic mapping as a visual aid to explore the relationships between different codes and to develop a coherent thematic structure [64]. This process involved an iterative approach of constant comparison, merging, and refining of themes until a satisfactory thematic map was generated.

Table 2 illustrates how the initial codes were organized into potential themes, along with descriptions of each theme.

We then carefully reviewed the potential themes to ensure their coherence and consistency with the coded data and the entire dataset. This phase involved two levels of reviewing: (1) checking if the coded data extracts within each theme formed a coherent pattern, and (2) examining if the thematic map accurately reflected the meanings evident in the dataset as a whole [65]. Through this process, some themes were refined, combined, or discarded to create a final thematic map that best captured the essence of the data.

Figure 2 presents the final thematic map, visually representing the main themes and their interrelationships in the context of generative AI-assisted foreign language teaching.

The central theme, “AI-Enhanced Teaching and Learning Practices”, encompasses three key sub-themes: “Transforming Pedagogical Approaches”, “Reconstructing Educational Roles”, and “Innovating Assessment Methods”. These sub-themes are closely interconnected and collectively contribute to the overarching theme of how GenAI reshapes foreign language education.

Surrounding the central theme are several other important themes that emerged from the data analysis. “Ethical and Policy Challenges” and “Cultural and Identity Issues” represent the broader societal and cultural implications of AI integration in education. “AI Literacy and Critical Thinking” and “Technology Adaptation and Development” highlight the new competencies and adaptations required from both educators and learners in the AI era. “Reshaping Educational Spaces and Resources” and “New Modes of Collaborative Learning” illustrate the transformative impact of AI on the physical and social dimensions of learning.

In the final phase of the thematic analysis, we aimed to produce a coherent and compelling narrative that tells the story of the data in relation to our research questions. We carefully selected vivid and representative data extracts to illustrate each theme and to support our interpretations. Following Tracy’s (2010) notion of “thick description”, we provided rich contextual details and in-depth explanations to enhance the credibility and resonance of our findings [66].

3.1.3. Research Findings

The thematic analysis of the rich qualitative data collected through classroom observations, in-depth interviews, document analyses, participatory observation, focus groups, and student reflective journals yielded a comprehensive understanding of the multifaceted impact of GenAI on foreign language teaching and learning. The findings are organized around the key themes identified in the analysis, which collectively paint a picture of the transformative potential, challenges, and implications of AI integration in educational contexts.

To illustrate the transformative changes brought about by AI integration, we present a detailed description of how AI was used in the three representative foreign language courses studied. In the English Literature course, AI language models were employed to generate multiple interpretations of classic literary works from various perspectives. For instance, when studying “Pride and Prejudice”, the AI-generated analyses from feminist, social class, and marriage viewpoints. As one student remarked, “The AI-generated interpretations opened my eyes to new ways of understanding the novel. It’s like having a panel of experts discussing the book from different angles”. (Student A, English Literature). This use of AI fundamentally transformed the way students engaged with literary texts, promoting critical thinking and multi-perspective analyses.

In the Chinese Language course, AI-powered adaptive learning systems were used to provide personalized feedback and recommendations based on individual student performance. The AI system continuously assessed students’ language proficiency levels and provided tailored content, exercises, and feedback to address their specific needs. As the teacher noted, “The AI system has revolutionized how we differentiate instruction. It’s like having a personal tutor for each student, adapting to their pace and style of learning”. (Teacher B, Chinese Language). This application of AI significantly enhanced the individualization and efficiency of language learning, transforming the traditional one-size-fits-all approach.

In the Spanish Culture course, AI-generated virtual reality (VR) simulations were used to immerse students in authentic cultural scenarios. Students engaged in virtual interactions with AI-powered characters representing different Spanish-speaking cultures, practicing their language skills and cultural knowledge in realistic contexts. As one student shared, “The AI-generated VR simulations made me feel like I was actually in Spain, talking to locals and experiencing the culture firsthand. It’s a game-changer for cultural immersion”. (Student C, Spanish Culture). This innovative use of AI transformed the way students learned about and experienced foreign cultures, bridging the gap between the classroom and the real world.

These examples demonstrate how educators have integrated AI technologies, such as language models, adaptive learning systems, and VR simulations, into their pedagogical approaches, leading to transformative changes in foreign language education. By designing personalized, immersive, and multi-perspective learning experiences with the help of AI, educators have reshaped the way students engage with language, the literature, and culture. As one professor summarized, “By incorporating AI into our teaching practices and learning design, we have fundamentally transformed the way we teach and learn foreign languages. It’s not just about using new tools; it’s about reimagining what’s possible in language education through the thoughtful application of AI”. (Professor D, Foreign Language Department).

Central to the findings is the theme of “AI-Enhanced Teaching and Learning Practices”, which encapsulates the fundamental changes in pedagogical approaches, educational roles, and assessment methods brought about by educators’ adoption and integration of GenAI. The data reveals that teachers have leveraged AI technologies, such as language models and adaptive learning systems, to significantly reshape the way foreign language content is delivered and interacted with. Teachers reported using AI-generated materials to provide students with more diverse, personalized, and engaging language input, as exemplified by one participant’s statement: “By incorporating the AI-generated multi-perspective interpretations of ‘Pride and Prejudice’, including feminist, social class, and marriage views, into my teaching, I was able to enrich and diversify the learning experience for my students”. (Observation notes, Week 3). This aligns with the notion of “AI-assisted language input” (AI-LI) discussed in recent literature, which highlights the potential of AI to support educators in enriching and contextualizing language learning experiences [5].

Central to the findings is the theme of “AI-Enhanced Teaching and Learning Practices”, which encapsulates the fundamental changes in pedagogical approaches, educational roles, and assessment methods brought about by educators’ adoption and integration of GenAI. The data reveals that teachers have leveraged AI technologies, such as language models and adaptive learning systems, to significantly reshape the way foreign language content is delivered and interacted with. Teachers reported leveraging AI-generated materials to provide students with more diverse, personalized, and engaging language input, as exemplified by one participant’s statement: “The AI-generated multi-perspective interpretations of ‘Pride and Prejudice’, including feminist, social class, and marriage views”. (Observation notes, Week 3). This aligns with the notion of “AI-assisted language input” (AI-LI) discussed in recent literature, which highlights the potential of AI to support educators in enriching and contextualizing language learning experiences [5].

Moreover, the findings suggest a profound shift in the roles and responsibilities of foreign language teachers in the AI era, driven by their strategic use of AI technologies. As one professor eloquently put it, “I now see myself more as a learning designer and facilitator, rather than just a transmitter of knowledge. It’s a paradigm shift in how we approach teaching”. (Professor W, English Literature). This resonates with the concept of “Teacher Perceived Role” (T-PR) in AI-integrated education, which emphasizes the transition from knowledge dispensers to learning architects and guides, enabled by the thoughtful application of AI [67]. The data also highlights the importance of “Teacher AI Literacy” (T-AI-L) and ongoing professional development (AI-TD) for educators to effectively harness the potential of AI in teaching practices [68].

From the student perspective, the findings reveal the impact of educators’ use of AI on transforming language learning experiences and outcomes. Students reported engaging in dynamic interactions with AI systems, receiving personalized feedback, and developing new learning strategies, all facilitated by their teachers’ integration of AI. As one student reflected, “AI has made me realize that the key isn’t memorizing information, but learning how to effectively utilize and analyze it. It’s changed how I approach my studies entirely”. (Student L, 3rd year). This echoes the notion of “Student Metalearning” (S-ML) in AI-supported environments, which involves learners’ active reflection on and regulation of their learning processes [69]. The data also underscores the critical role of “Student AI Literacy” (S-AI-L) in navigating the complexities and challenges of AI-mediated learning, which educators can cultivate through their teaching practices.

Beyond the immediate classroom context, the findings illuminate the broader ethical, social, and cultural implications of educators’ AI integration in foreign language education. Participants raised concerns about issues such as academic integrity, data privacy, and algorithmic biases, as encapsulated in the theme of “Ethical and Policy Challenges”. As one department head noted, “Ensuring academic integrity in an AI-assisted environment is one of our biggest challenges. We’re still grappling with where to draw the line”. (Professor L, Department Head). This resonates with the growing body of literature on the ethical dimensions of AI in education, which calls for proactive policies and guidelines to mitigate potential risks and ensure responsible use of AI technologies [70].

The findings also highlight the cultural and identity-related implications of educators’ use of AI in foreign language learning. Participants discussed how AI technologies might influence learners’ cultural understanding, intercultural competence, and sense of self as language learners. As one student poignantly reflected, “Using AI has made me question what it means to be ‘fluent’ or ‘proficient’ in a language. It’s reshaping how I view my language learning journey”. (Student W, 4th year). This aligns with recent discussions on the complex interplay between AI, language, culture, and identity in educational contexts, and the role of educators in navigating these dynamics [71].

Furthermore, the findings underscore the impact of educators’ AI integration on transforming the physical and social dimensions of foreign language learning. Participants reported changes in the use of educational spaces and resources, as well as the emergence of new modes of collaborative learning, all driven by teachers’ incorporation of AI. As one researcher observed, “Traditional classroom layouts prove challenging for AI-assisted group collaborations. There’s a growing need for more flexible, technology-integrated space designs”. (Researcher’s reflective journal, Week 4). This resonates with the concept of “Spatial Design Impact” (SP-DI) in AI-integrated learning environments, which emphasizes the importance of designing spaces that facilitate seamless human-AI interactions and collaborations, in line with educators’ pedagogical goals [72].

Lastly, the findings shed light on the process and challenges of technology adaptation and development in the context of GenAI integration. Participants reported a learning curve effect, with initial struggles gradually giving way to improved efficiency and effectiveness over time, as they refined their pedagogical approaches. As one researcher noted, “Observed a clear learning curve effect. Initial weeks showed struggles with AI integration, but by mid-semester, both teachers and students demonstrated markedly improved efficiency”. (Researcher’s reflective journal, Week 10). This aligns with the notion of “Technology Adaptation” (TC-AD) in educational innovation, which emphasizes the iterative nature of technology integration and the importance of ongoing support and training for educators [73].

The thematic analysis of the qualitative data yielded a rich tapestry of findings that illuminate the multifaceted impact of GenAI on foreign language teaching and learning. The findings underscore the transformative potential of AI in reshaping pedagogical practices, learning experiences, and educational roles, while also highlighting the ethical, social, and cultural implications of AI integration. The insights generated from this study provide a solid empirical foundation for further theoretical development and practical recommendations in the field of AI-assisted language education. However, this does not provide a comprehensive and more nuanced understanding of the intrinsic mechanistic relationship between the interactions of technology, pedagogy, and learning, and qualitative research cannot accurately quantify how a change in one of the variables affects the other observed variables; therefore, we supplemented and enriched the research design of this paper by further designing a simulation and modeling study in the SD-ABM paradigm based on the framework that we have proposed based on the qualitative study.

3.2. Modeling Simulation: The Use of GenAI in Foreign Language Education Systems and Its Implications

3.2.1. Simulation Modeling Process

Based on the GAIFL-STS model and our qualitative research findings, we designed a complex hybrid simulation model combining system dynamics (SD) and agent-based modeling (ABM) methods to comprehensively simulate the application and impact of GenAI in foreign language education systems. The hybrid simulation model integrates two complementary approaches: system dynamics (SD) and agent-based modeling (ABM). SD is a top-down approach that focuses on the macro-level behavior of the system, capturing the complex feedback loops, accumulations, and delays among key variables. It represents the system as a set of stocks (e.g., AI technology maturity, teacher AI literacy), flows (e.g., technology development rate, teacher training rate), and auxiliary variables (e.g., institutional support and ethical regulations) that interact dynamically over time. ABM, on the other hand, is a bottom-up approach that simulates the micro-level interactions and adaptations of heterogeneous agents (e.g., teachers, students, AI systems) based on their individual characteristics, decision rules, and learning mechanisms. It captures the emergent properties and unintended consequences that arise from the local interactions among agents and their environment.

By combining SD and ABM, our hybrid simulation model provides a multi-level and multi-perspective understanding of the complex dynamics and impacts of GenAI integration in foreign language education. The SD component captures the high-level system behavior and performance, such as the diffusion of AI technologies, the enhancement of teaching and learning practices, and the evolution of institutional and ethical frameworks. The ABM component, nested within the SD structure, simulates the individual and social responses of teachers, students, and AI systems to the changing technological, pedagogical, and organizational contexts. The integration of SD and ABM enables a more comprehensive and realistic representation of the GAIFL-STS, accounting for both the structural dynamics and the actor agency in shaping the trajectories and outcomes of AI-assisted language learning.

(1) System Dynamics (SD) Component

a. Defining System Boundaries

We expanded the system boundaries to incorporate more granular variables that capture the multifaceted nature of AI-assisted language education. These variables encompassed key dimensions such as AI technology characteristics (e.g., language generation capability, context understanding, personalized recommendation), teacher factors (e.g., AI literacy, technology acceptance, professional development), student factors (e.g., language proficiency, learning motivation, AI-assisted learning acceptance), organizational factors (e.g., AI infrastructure investment, policy support, innovation culture), teaching process (e.g., AI-assisted teaching ratio, personalized learning path diversity), learning outcomes (e.g., language skill improvement, critical thinking, AI literacy), and ethical and social impact (e.g., data privacy protection, educational equity, cultural diversity). The inclusion of these variables aligned with the multidimensional nature of the GAIFL-STS model, capturing the complex interplay among technology, pedagogy, learners, and context in AI-enhanced language education [1,2]. The expanded system boundaries enable a more holistic and nuanced examination of the dynamic processes and emergent properties of the system.

b. Constructing Causal Loop Diagrams

We identified several key feedback loops in the system, including both positive and negative loops:

Positive Loops:

• (1). AI Technology Enhancement Loop: AI language generation capability→Learning resource richness→Student learning motivation→Learning outcomes→Social acceptance of AI education→AI infrastructure investment→AI language generation capability

• (2). Teacher Professional Development Loop: AI literacy level→AI-assisted teaching time ratio→Teaching innovation→Student learning satisfaction→Teacher job satisfaction→Professional development investment→AI literacy level

• (3). Organizational Learning Loop: AI infrastructure investment→AI-assisted teaching effectiveness→Learning outcomes→Organizational innovation culture→Cross-departmental collaboration→AI infrastructure investment

Negative Loops:

• (4). Ethical Challenge Loop: AI usage level→Data privacy risk→Ethical norms completeness→AI usage restrictions→AI usage level

• (5). Role Conflict Loop: AI-assisted teaching time ratio→Teacher role change→Role identity decline→Teacher resistance→AI-assisted teaching time ratio

• (6). Skill Substitution Loop: AI capability enhancement→Certain language skill demand decrease→Student learning motivation change→Traditional language skill level→Dependence on AI→AI capability enhancement

The identification of these feedback loops was crucial for understanding the nonlinear dynamics and unintended consequences of AI integration in education systems. The positive loops represented self-reinforcing processes that can lead to virtuous cycles of technology enhancement, teacher development, and organizational learning. In contrast, the negative loops illustrated potential balancing processes and undesirable outcomes, such as ethical dilemmas, role conflicts, and skill substitution.

c. Developing Stock and Flow Diagrams

We designed sophisticated mathematical equations to describe the relationships among the key variables in the system. These equations captured the accumulation and depletion processes of stocks, such as AI language generation capability, teacher AI literacy, student language ability, and organizational support. The rates of change were determined by a combination of endogenous and exogenous factors, reflecting the complex interdependencies within the system. For example:

ai_language_generation(t) = ai_language_generation(t-dt) + (tech_progress_rate − tech_obsolescence_rate) ∗ dt tech_progress_rate = f(organizational_support, global_AI_RD_trends) tech_obsolescence_rate = constant ∗ ai_language_generation

teacher_ai_literacy(t) = teacher_ai_literacy(t-dt) + (literacy_improvement_rate − literacy_decay_rate) ∗ dt literacy_improvement_rate = f(professional_development_investment, teaching_ai_time, organizational_training_intensity) literacy_decay_rate = constant ∗ teacher_ai_literacy

These equations were implemented in the sd_derivatives method of our AIEducationModel class, ensuring a dynamic and interconnected representation of the system’s behavior over time. The equations captured the accumulation and depletion processes of key variables, such as AI language generation capability, teacher AI literacy, student language proficiency, organizational innovation culture, and ethical norms completeness. The rates of change were determined by a combination of endogenous and exogenous factors, reflecting the complex interdependencies within the system.

(2) Agent-Based Modeling (ABM) Component

The ABM component complemented the SD model by providing a bottom-up perspective on the microlevel interactions and emergent behaviors of individual agents. We designed three types of agents: TeacherAgent, StudentAgent, and AISystemAgent, each with specific attributes and behavioral rules.

a. TeacherAgent: Attributes: ai_literacy, work_satisfaction, innovation_tendency, role_identity Behavioral rules:

• Update AI literacy based on AI language generation capability and organizational support

• Adjust work satisfaction based on organizational support and student language ability

• Modify innovation tendency based on AI language generation and organizational innovation

• Adapt role identity based on innovation tendency and AI-assisted teaching time

b. StudentAgent: Attributes: language_ability, learning_motivation, ai_acceptance, self_learning_ability, cross_cultural_sensitivity Behavioral rules:

• Improve language ability based on AI language generation and teacher AI literacy

• Adjust learning motivation based on social acceptance and personal language ability

• Update AI acceptance based on AI language generation and social acceptance

• Enhance self-learning ability based on AI acceptance and learning resource richness

• Develop cross-cultural sensitivity based on cross-cultural opportunities

c. AISystemAgent: Attributes: language_model_complexity, personalization_accuracy, response_speed, error_rate, learning_ability Behavioral rules:

• Enhance language model complexity based on global AI advancements

• Improve personalization accuracy through continuous learning

• Increase response speed with technological improvements

• Reduce error rate through system optimizations

• Adapt learning ability based on interactions with teachers and students

The environment design incorporated the broader educational context, including organizational factors, ethical considerations, and social dynamics. Interaction rules were implemented through the influence of agent behaviors on the system-level variables and vice versa.

(3) Effective Integration of SD and ABM

To achieve an effective integration of SD and ABM, we employed several strategies:

1. Multiscale modeling: The SD model captures macro-level system dynamics, while the ABM simulates micro-level individual behaviors. We defined clear interface variables to allow interaction between the two scales.

2. Time synchronization: Both SD and ABM components use the same time step, ensuring consistent temporal evolution of the system.

3. Data flow: ① Bottom-up: Aggregated agent behaviors influence SD variables (e.g., average teacher AI literacy affecting the system-level teacher_ai_literacy variable). ② Top-down: SD variables set the environment parameters for agent behaviors (e.g., ai_language_generation influencing individual agents’ attributes and decisions).

4. Feedback loops: Cross-model feedback loops are implemented, such as the system-level social_acceptance influencing individual agents’ AI acceptance, which in turn affects the system-level variable.

5. Parameter calibration: We use ABM micro-data to calibrate SD parameters and vice versa, enhancing the overall model accuracy and reliability.

6. Emergent behavior detection: The ABM component allows for the detection of emergent behaviors, which can then be incorporated into the SD model as new variables or relationships.

7. Scenario analysis: We design comprehensive scenarios that span both SD and ABM components, exploring how new AI policies might simultaneously affect system dynamics and individual behaviors.

The integration of SD and ABM is a powerful approach for modeling complex adaptive systems like AI-assisted language education. By combining the strengths of both methods, the hybrid model can provide a more comprehensive and realistic representation of the dynamic processes, feedback mechanisms, and emergent properties at multiple levels of the system. This model provides a flexible and extensible platform for scenario analyses and policy experimentation. By manipulating system-level parameters in SD and agent-level rules in ABM, researchers and decision-makers can explore a wide range of “what-if” scenarios and assess the effectiveness of different interventions in a virtual environment before implementing them in the real world. This enables a proactive and evidence-based approach to the design and management of AI-assisted language education systems.

In summary, the hybrid simulation modeling process, combining SD and ABM, offers a powerful tool for understanding and managing the complex dynamics of GenAI integration in foreign language education. The expanded system boundaries, causal loop diagrams, stock and flow equations, agent attributes, behavioral rules, and interaction patterns provide a rich and nuanced representation of the multifaceted system. The effective integration strategies ensure the consistency and validity of the model, while enabling a holistic exploration of the emergent properties and policy implications. This modeling process lays a solid foundation for the subsequent simulation experiments and analyses.

3.2.2. Model Simulation Results

The hybrid SD-ABM simulation model was run under various scenarios to investigate the complex dynamics and emergent properties of GenAI integration in foreign language education. The simulation results not only validated the effectiveness of the GAIFL-STS model but also revealed the intricate interactions between GenAI and various elements of the foreign language education system. The following sections present a detailed analysis of the simulation results from both system-level and individual-level perspectives.

At the system level, the evolution trends in SD variables (as shown in Figure 3) unveil the global impact of GenAI on the foreign language education system. Firstly, the AI language generation capability (ai_language_generation) exhibits an S-shaped curve, characterized by rapid growth followed by stabilization. This pattern aligns with the theory of innovation diffusion, reflecting the process of initial adoption, rapid diffusion, and eventual saturation of GenAI in foreign language education. Notably, the growth of AI language generation capability is positively correlated with teacher AI literacy (teacher_ai_literacy) and student language ability (student_language_ability), but with varying growth rates and magnitudes. This heterogeneous growth pattern resonates with the Technological Pedagogical Content Knowledge (TPACK) framework, suggesting that the integration of technology, pedagogy, and subject content knowledge is a complex and nonlinear process.

Secondly, the organizational support (organizational_support) and innovation culture (organizational_innovation) variables exhibit a fluctuating upward trend, reflecting the iterative trial-and-error and continuous adjustment process that educational institutions undergo when adapting to AI technologies. This finding is consistent with the organizational learning theory, highlighting the dynamic adaptability of organizations in the technology innovation adoption process. In particular, we observed a positive feedback loop between organizational support and teacher satisfaction (teacher_satisfaction), underscoring the pivotal role of institutional support in promoting teachers’ acceptance and effective use of AI technologies.

Regarding ethical and social impact, the simulation results show that ethical standards (ethical_standards) and data privacy (data_privacy) variables initially decline during the rapid development of AI technology but gradually recover and stabilize over time. This dynamic process reflects the “lag effect” of technology ethics in the innovation adoption process. Interestingly, the social acceptance (social_acceptance) variable follows a fluctuating upward trend, indicating that public attitudes towards AI-assisted foreign language education evolve from initial skepticism to gradual acceptance, aligning with the process of social system acceptance of new technologies described in innovation diffusion theory.

At the individual level, the agent-based modeling results (as shown in Figure 4, Figure 5 and Figure 6) provide more microscopic and dynamic insights. The attribute distribution of TeacherAgent (as shown in Figure 4) reveals that AI Literacy and Innovation Tendency exhibit positively skewed distributions, indicating that most teachers are at relatively low levels during the early stages of AI application, with a small proportion of early adopters. This distribution characteristic is consistent with the proportion of innovators and early adopters described in innovation diffusion theory. Concurrently, we observed a bimodal distribution of Work Satisfaction and Role Identity, reflecting the polarized reactions of the teacher population when facing AI technology. This finding supports Ertmer and Ottenbreit-Leftwich’s (2010) discussion of intrinsic and extrinsic barriers in teachers’ technology integration process [74], suggesting that teacher professional development and support strategies need to adopt differentiated approaches for different groups.

The attribute distribution of StudentAgent (as shown in Figure 5) presents a different pattern. Language Ability and Learning Motivation exhibit approximately normal distributions, with the peak of the distribution shifting to the right over time, indicating that AI-assisted learning has a positive impact on the language ability and learning motivation of most students. Notably, AI Acceptance and Self-learning Ability show a significant positive correlation, supporting Lai and Gu’s (2011) view that technology-enhanced language learning promotes learner autonomy [75]. However, we also observed a high dispersion in the distribution of Cross-cultural Sensitivity, prompting us to consider how to more effectively cultivate students’ cross-cultural competence while promoting language learning through AI.

The attribute distribution of AISystemAgent (as shown in Figure 6) reflects the evolutionary characteristics of AI technology itself. Language Model Complexity and Personalization Accuracy show a continuous growth trend, consistent with the rapid advancement of deep learning techniques. However, the improvement rates of Response Speed and Error Rate are relatively slower, reflecting the challenges that AI systems still face in terms of real-time interaction and reliability. This finding resonates with a discussion of the “performance-usability trade-off” in AI educational applications, suggesting that while pursuing AI performance enhancement, we need to place equal emphasis on its applicability and reliability in educational scenarios.

The correlation heatmap (Figure 7) further reveals the complex interactions among system variables. We observed strong positive correlations between AI language generation capability and student language ability, teacher AI literacy, and organizational innovation culture, validating the hypothesis of mutual reinforcement among technological, pedagogical, and organizational factors in our GAIFL-STS model. However, AI language generation capability exhibits weak negative correlations with ethical standards and data privacy, again highlighting the potential conflict between technological progress and ethical considerations, necessitating the simultaneous strengthening of ethical frameworks while advancing AI educational applications.

Finally, the sensitivity analysis results (Figure 8) provide insights into the dynamic impact of key system variables. We found that the initial AI language generation capability and teacher AI literacy have the most significant long-term impact on student language ability, emphasizing the importance of the coordinated development of technological capability and teacher literacy in AI-assisted foreign language education. Meanwhile, the impact of organizational support on the system’s long-term performance exhibits nonlinear characteristics, suggesting the existence of an optimal level of support, with both excessive and insufficient support potentially affecting system effectiveness.

Overall, the simulation results of this study not only validate the effectiveness of the GAIFL-STS model but also reveal the complex dynamics of GenAI application in the foreign language education system. These findings provide a systematic and multi-level perspective for understanding AI-driven educational transformation while offering valuable insights for policy-making and educational practice. However, we also recognize that the interpretation of simulation results needs to be combined with specific educational contexts and cultural backgrounds, and future research should further explore the applicability and generalizability of the model in different educational systems.

3.3. Integration Analysis: Complementary Theoretical Findings Between Qualitative Research Findings and Model Simulation Results

The integration of qualitative research findings and model simulation results provides a comprehensive and synergistic understanding of the complex dynamics and impacts of GenAI in foreign language education. The qualitative findings offer rich, contextualized insights into the perceptions, experiences, and practices of key stakeholders, while the simulation results reveal the systemic patterns, emergent properties, and potential trajectories of AI integration. By juxtaposing and synthesizing these two strands of findings, we can derive complementary theoretical insights that enhance the explanatory power and practical relevance of the GAIFL-STS model.

First, the qualitative findings provide a nuanced and grounded understanding of the multifaceted affordances and challenges of GenAI in foreign language teaching and learning. The thematic analysis unveiled the transformative potential of AI in enhancing language input, output, and interaction, as well as in enabling personalized feedback and adaptive learning. These findings resonate with the simulation results, which demonstrate the positive impact of AI language generation capability on student language proficiency and learning motivation. The qualitative accounts also revealed the crucial role of teacher AI literacy and innovative pedagogy in mediating the effective integration of AI technologies, corroborating the simulation findings on the importance of teacher professional development and the co-evolution of technological and pedagogical factors.

Moreover, the qualitative findings shed light on the complex interplay between AI affordances and human agency in shaping the trajectories and outcomes of AI-assisted language education. The interviews and observations surfaced the agentic responses of teachers and students in negotiating, appropriating, and resisting AI technologies, highlighting the non-deterministic and context-dependent nature of AI integration. These insights complement the simulation results, which capture the emergent patterns and unintended consequences arising from the micro-level interactions among agents with diverse attributes and behavioral rules. The integration of qualitative and simulation findings thus offers a more holistic and realistic understanding of the socio-technical dynamics of AI in education, eschewing both technological determinism and social reductionism.

Second, the qualitative findings illuminate the ethical, social, and cultural implications of GenAI in foreign language education, which are not fully captured by the simulation model. The thematic analysis revealed the concerns and tensions around issues such as data privacy, algorithmic bias, academic integrity, and cultural diversity, underscoring the importance of ethical governance and critical reflection in the adoption of AI technologies. These findings provide a necessary complement to the simulation results, which focus primarily on the technical and organizational dimensions of AI integration. The qualitative insights also highlight the situated and value-laden nature of AI adoption, reminding us to attend to the broader social and cultural contexts that shape the meanings and consequences of AI in education.

Third, the simulation results offer a systemic and dynamic perspective on the long-term trajectories and policy implications of GenAI in foreign language education, extending the qualitative findings. The scenario analysis and sensitivity tests revealed the complex feedback loops, tipping points, and trade-offs involved in AI integration, such as the tension between personalization and social interaction, the trade-off between AI performance and usability, and the optimal level of organizational support. These findings provide a forward-looking and strategic lens for interpreting the qualitative insights, informing the design and management of AI-assisted language education systems. The simulation results also underscore the importance of a holistic and adaptive approach to AI governance, attending to the interplay of technological, pedagogical, organizational, and ethical factors over time.

Fourth, the integration of qualitative and simulation findings enables a more robust and trustworthy theory-building process. The qualitative findings ground the conceptual model in the lived realities and situated knowledge of educational stakeholders, ensuring its ecological validity and practical relevance. The simulation results, in turn, provide a rigorous and systematic way to test, refine, and extend the theoretical propositions derived from the qualitative analysis, enhancing their internal consistency and logical coherence. The iterative dialog between qualitative and simulation insights allows for a more abductive and pragmatic approach to theory development, combining the strengths of inductive exploration and deductive verification.

Finally, the complementary findings from qualitative research and model simulation point to promising directions for future research and practice. The qualitative insights suggest the need for more participatory and design-based research approaches that actively involve teachers and students in the co-creation and appropriation of AI technologies, fostering a more empowering and inclusive vision of AI in education. The simulation results highlight the potential of agent-based modeling as a powerful tool for exploring the complex dynamics and emergent properties of AI-assisted learning systems, informing the design and testing of adaptive and resilient governance strategies. The integration of qualitative and simulation methods also calls for more interdisciplinary and mixed-methods research that bridges the divide between the social and computational sciences, fostering a more holistic and actionable understanding of AI in education.

4. Discussion

4.1. Summary of Research Findings

This study aimed to investigate the complex dynamics and impacts of GenAI integration in foreign language education through the lens of the GAIFL-STS model. Grounded in socio-technical systems theory and the TPACK framework, the GAIFL-STS model provides a holistic and multi-level perspective on the interplay of technological, pedagogical, organizational, and ethical factors in shaping the trajectories and outcomes of AI-assisted language learning.

To empirically validate and refine the conceptual model, we adopted a mixed-methods approach that combines qualitative case studies and hybrid simulation modeling. The qualitative phase involved in-depth interviews, participant observations, and document analyses to explore the perceptions, experiences, and practices of key stakeholders in three representative foreign language courses at a Chinese university. This phase directly addressed the first research question by providing rich insights into the affordances and challenges of GenAI technologies for foreign language teaching and learning, as perceived by teachers and learners. The thematic analysis of the qualitative data revealed that GenAI technologies, such as language models and adaptive learning systems, have the potential to enhance language input, output, and interaction, as well as to enable personalized feedback and adaptive learning. Teachers and learners perceived AI as a powerful tool for providing authentic and diverse language exposure, facilitating immersive and interactive learning experiences, and promoting learner autonomy and engagement. However, the findings also highlighted the challenges and concerns associated with AI adoption, such as data privacy, algorithmic bias, academic integrity, and cultural diversity.

The simulation phase employed a novel integration of system dynamics (SD) and agent-based modeling (ABM) to capture the complex feedback loops, emergent properties, and unintended consequences of AI integration. This phase directly addressed the third research question by examining the extent to which, and under what conditions, GenAI can enhance the effectiveness, efficiency, and equity of foreign language education, as well as the potential unintended consequences and ethical implications. The SD model depicted the macro-level system behavior and the dynamic interactions among key variables, such as AI language generation capability, teacher AI literacy, student language proficiency, organizational innovation culture, and ethical norms completeness. The ABM model represented the micro-level actions and interactions of heterogeneous agents (teachers, students, and AI systems) with diverse attributes, behavioral rules, and learning mechanisms. The hybrid simulation experiments and sensitivity analyses revealed the nonlinear dynamics, tipping points, and trade-offs involved in AI integration, as well as the long-term trajectories and policy implications. The results indicated that GenAI has the potential to enhance the effectiveness, efficiency, and equity of foreign language education under certain conditions, such as high levels of teacher AI literacy, student motivation, and the quality of human-AI interaction. However, the simulation also revealed the potential risks and unintended consequences of AI adoption, such as the widening of achievement gaps between advantaged and disadvantaged students, the erosion of teacher autonomy and creativity, and the homogenization of language and culture.

In response to the three research questions proposed in the introduction, the study findings provide the following comprehensive insights:

Regarding the affordances and challenges of GenAI technologies for foreign language teaching and learning (RQ1), the qualitative findings revealed the multifaceted affordances of GenAI in enhancing language input, output, and interaction, as well as in enabling personalized feedback and adaptive learning. AI technologies were perceived by teachers and learners as powerful tools for enriching the linguistic environment, providing authentic and diverse language exposure, and facilitating immersive and interactive learning experiences. Participants highlighted the potential of AI in promoting learner autonomy, engagement, and motivation by offering personalized content recommendations, adaptive assessments, and timely feedback. However, the findings also highlighted the challenges and concerns associated with AI adoption, such as the issues of data privacy, algorithmic bias, academic integrity, and cultural diversity. Teachers expressed apprehension about the potential misuse of AI for cheating and plagiarism, while students raised questions about the fairness and transparency of AI-based assessments. Both groups emphasized the importance of developing AI literacy and critical thinking skills to navigate the complexities and uncertainties of AI-assisted learning. These findings provide a nuanced and balanced understanding of the affordances and challenges of GenAI in language education, underscoring the need for proactive strategies and policies to maximize the benefits and mitigate the risks of AI adoption.

Concerning the adaptation of educational institutions’ strategies, policies, and practices to support the integration of GenAI in foreign language programs (RQ2), the qualitative findings demonstrated the critical role of educational institutions in shaping the ways AI technologies are leveraged to enhance language teaching and learning. The findings revealed that by designing personalized, immersive, and multi-perspective learning experiences with the help of AI, educators have reshaped the way students engage with language, the literature, and culture. Institutions that successfully integrated AI technologies adopted a holistic approach that focused on empowering teachers to effectively incorporate AI into their pedagogical practices. This involved providing professional development opportunities for teachers to enhance their AI literacy and skills in designing AI-assisted learning activities. Moreover, institutions supported teachers in leveraging AI technologies to reshape the way foreign language content is delivered and interacted with, such as using AI-generated materials to provide students with diverse, personalized, and engaging language input. These policies were developed through collaborative efforts involving teachers, administrators, and AI developers to ensure that AI integration aligns with the pedagogical goals and values of the institution.

With respect to the extent to which and under what conditions GenAI can enhance the effectiveness, efficiency, and equity of foreign language education (RQ3), the simulation experiments and sensitivity analyses indicated that GenAI has the potential to enhance the effectiveness, efficiency, and equity of foreign language education under certain conditions. The scenario analysis showed that the positive impact of AI on student learning outcomes was contingent upon factors such as teacher AI literacy, student motivation, and the quality of human-AI interaction. Institutions that invested in teacher professional development, student digital literacy training, and user-centered AI design were more likely to reap the benefits of AI for personalized and adaptive learning. However, the simulation also revealed the potential risks and unintended consequences of AI adoption, such as the widening of achievement gaps, the erosion of teacher autonomy, and the homogenization of language and culture. The sensitivity analysis further highlighted the critical importance of ethical governance and responsible innovation in mitigating these risks and ensuring the socially beneficial and inclusive development of AI in education. These findings provide a systemic and dynamic understanding of the conditions and factors that shape the impact of AI on language learning, informing the design of evidence-based policies and interventions to optimize the outcomes and mitigate the risks of AI adoption.

The integration of qualitative and simulation findings yielded complementary and corroborative insights that enhanced the explanatory power and practical relevance of the GAIFL-STS model. The qualitative findings provided a grounded and contextualized understanding of the lived experiences and situated practices of educational stakeholders, while the simulation results offered a systemic and predictive perspective on the long-term patterns and potential outcomes of AI integration. The iterative dialog between these two strands of findings enabled a more robust and trustworthy theory-building process, informing the design and governance of AI-assisted language education systems.

4.2. Limitations and Future Research

While this study makes significant contributions to the understanding of AI integration in foreign language education, it also has several limitations that should be acknowledged and addressed in future research.

First, although the qualitative findings offer rich and contextual insights, they are based on a relatively small sample of cases from a single university context, which may limit their transferability to other educational settings. Qualitative research does not aim for statistical generalizability, but rather for in-depth understanding and analytical transferability. However, the limited diversity in the sample, both in terms of the number of participants and the institutional context, may have introduced certain biases or overlooked important variations in experiences and perspectives. Future studies could explore the applicability and resonance of the findings in diverse institutional and cultural contexts, involving a larger and more heterogeneous sample of participants from multiple universities and regions.

Second, the simulation model, while grounded in empirical data and theoretical frameworks, is still a simplified representation of the complex reality of AI-assisted language learning. The model’s assumptions, parameters, and relationships are based on the limited data available and the researchers’ interpretations, which may not fully capture the intricacies and nuances of the real-world system. Additionally, the model’s sensitivity to initial conditions and parameter values may affect its robustness and reliability. Future research could refine and validate the model with more extensive data and stakeholder input, incorporate additional variables and mechanisms to enhance its fidelity and predictive power, and conduct more rigorous sensitivity analyses and uncertainty assessments to establish the boundaries and limitations of the model’s applicability.

Third, the study did not directly measure the long-term learning outcomes and social impacts of AI adoption, which require longitudinal and large-scale research designs. The cross-sectional nature of the qualitative data and the relatively short time frame of the simulation experiments may not adequately capture the dynamic and evolving nature of AI integration and its effects on student learning, motivation, and well-being over time. Moreover, the study’s focus on the immediate classroom and institutional contexts may have overlooked the broader societal implications and unintended consequences of AI adoption, such as issues of educational equity, digital divide, and algorithmic bias. Future studies could employ quasi-experimental or randomized controlled trials to rigorously assess the effects of AI interventions on student achievement, motivation, and well-being, as well as their broader implications for educational equity and social justice, using longitudinal designs and mixed-methods approaches.

Furthermore, the study’s reliance on self-reported data from interviews and surveys may have introduced certain response biases or social desirability effects, as participants may have provided answers that they believed were expected or socially acceptable, rather than their authentic experiences and opinions. The use of multiple data sources and triangulation techniques, such as classroom observations and learning analytics, could help mitigate these biases and provide a more comprehensive and objective understanding of the phenomenon under study.

Despite these limitations, this study provides a timely and critical exploration of the promises, perils, and possibilities of AI-assisted language learning and offers a roadmap for future research and innovation in this rapidly evolving field. It calls for a holistic, contextual, and ethical approach to the design, implementation, and evaluation of AI-assisted language learning systems, and highlights the importance of interdisciplinary and participatory research that engages multiple stakeholders in the co-creation of educational futures. Future research could build upon the findings and insights of this study, addressing its limitations and expanding its scope to advance the theoretical understanding and practical application of AI in language education and beyond.

5. Conclusions

5.1. Practical Implications

Despite these limitations, the study offers several practical implications for the design, implementation, and governance of AI-assisted language education systems. First, the findings suggest that language teachers and learners should be actively involved in the design and appropriation of AI technologies, rather than being treated as passive users or recipients. This requires the development of participatory and user-centered design approaches that leverage the situated knowledge and agentic capacities of educational stakeholders.

Second, the study highlights the importance of providing comprehensive and ongoing professional development for language teachers to enhance their AI literacy and pedagogical innovation. This includes not only technical training on AI tools and platforms but also critical reflection on the ethical, social, and cultural implications of AI adoption. Language teachers should be empowered to adapt and integrate AI technologies in ways that align with their pedagogical beliefs, contextual needs, and student diversity.

Third, the findings underscore the need for educational institutions to develop holistic and adaptive strategies for AI integration that attend to the interplay of technological, pedagogical, organizational, and ethical factors. This requires the establishment of cross-functional teams and collaborative partnerships that bring together expertise from different domains, such as language education, AI engineering, learning analytics, and educational policy. Institutions should also regularly monitor and evaluate the impact of AI adoption on student learning, teacher well-being, and organizational culture, and make timely adjustments based on the feedback and evidence.

Fourth, the study calls for the development of ethical guidelines and governance frameworks for the responsible innovation and use of AI in language education. This includes the establishment of data privacy and security protocols, the transparency and accountability of AI algorithms, and the protection of student agency and diversity. Educational institutions should also foster open dialogs and public engagement on the social and ethical implications of AI in education, involving multiple stakeholders such as students, parents, educators, researchers, policymakers, and industry partners.

Finally, the study suggests that the successful integration of GenAI in foreign language education requires a culture of openness, collaboration, and continuous learning. Educational stakeholders should be encouraged to experiment with new pedagogical approaches, share best practices, and learn from failures. Institutions should also cultivate a culture of evidence-based decision-making and reflective practice, using learning analytics and feedback loops to inform the iterative improvement of AI-assisted language learning systems.

5.2. Theoretical Contributions

This study makes several important theoretical contributions to the field of AI-assisted language education and the broader discourse on educational technology innovation. First, the GAIFL-STS model advances a holistic and integrative framework for understanding the complex socio-technical dynamics of AI in education. By synthesizing insights from socio-technical systems theory, the TPACK framework, and complex adaptive systems theory, the model captures the multi-level, multi-dimensional, and multi-stakeholder nature of AI integration, eschewing both technological determinism and social reductionism. The model highlights the critical role of human agency, social context, and ethical considerations in shaping the meanings and consequences of AI adoption, offering a more nuanced and realistic perspective than techno-centric or instrumental approaches.

Second, the study contributes to the growing body of literature on the affordances and challenges of GenAI in language education. The qualitative findings unveil the transformative potential of AI in enhancing language input, output, and interaction, as well as in enabling personalized feedback and adaptive learning. At the same time, the findings highlight the ethical, social, and cultural tensions and unintended consequences of AI adoption, such as the issues of data privacy, algorithmic bias, academic integrity, and cultural diversity. These findings provide a more balanced and critical perspective on the promises and perils of AI in education, informing the responsible design and governance of AI-assisted language learning systems.

Third, the study advances the theoretical understanding of the complex interplay between technological affordances and human agency in educational innovation. The qualitative findings reveal the agentic responses of teachers and students in negotiating, appropriating, and resisting AI technologies, highlighting the non-deterministic and context-dependent nature of AI integration. The simulation results, in turn, capture the emergent patterns and unintended consequences arising from the micro-level interactions among agents with diverse attributes and behavioral rules. These findings challenge the simplistic notions of technology as a neutral tool or an autonomous force, emphasizing the co-constitutive relationship between technology and human practices in educational change.

Finally, the study contributes to the conceptual and methodological development of socio-technical systems research in education. The GAIFL-STS model provides a generic and adaptable framework that can be applied and extended to other educational contexts and technologies beyond foreign language learning and GenAI. The hybrid SD-ABM simulation approach offers a powerful and flexible tool for modeling and analyzing the complex dynamics and emergent properties of educational socio-technical systems, informing the design and testing of adaptive and resilient governance strategies. The study also demonstrates the value of interdisciplinary and mixed-methods research that bridges the divide between the social and computational sciences, fostering a more holistic and actionable understanding of educational innovation.

Author Contributions

Conceptualization, Y.Z.; methodology, C.D.; investigation and resources, Y.Z.; validation, Y.Z. and C.D.; formal analysis, Y.Z. and C.D.; data curation, C.D.; writing—original draft, Y.Z. and C.D.; writing—review and editing, Y.Z.; visualization, C.D.; supervision, Y.Z.; funding acquisition, Y.Z. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

All methods followed the guidelines and regulations of the 1964 Helsinki Declaration and its later amendments. Since this study does not involve intervention and is low-risk, ethical review and approval were waived according to the institutional review boards at the Faculty of Humanities and Social Sciences at Harbin Institute of Technology.

Informed Consent Statement

Informed consent was obtained from all subjects involved in the study.

Data Availability Statement

The article encompasses the study’s original contributions, further inquiries can be addressed to the corresponding author.

Conflicts of Interest

The authors declare no conflicts of interest.

Footnotes

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Figures and Tables

Figure 1. Generative AI-assisted Foreign Language Education Socio-Technical System (GAIFL-STS) model.

Figure 2. Thematic map of generative AI-assisted foreign language teaching.

Figure 3. Trends in SD variables for the global impact of GenAI on the foreign language education system.

Figure 4. The agent-based modeling results: attribute distribution of TeacherAgent.

Figure 5. The agent-based modeling results: attribute distribution of StudentAgent.

Figure 6. The agent-based modeling results: attribute distribution of AISystemAgent.

Figure 7. The correlation heatmap among system variables.

Figure 8. The results of sensitivity analysis of key system variables.

Table 1

Initial coding example for generative AI-assisted foreign language teaching.

Data Source	Initial Code	Code Description	Example Quotes
Classroom Observation	AI-LI (AI-assisted Language Input)	AI system providing language input materials	“The AI-generated multi-perspective interpretations of ‘Pride and Prejudice’, including feminist, social class, and marriage views”. (Observation notes, Week 3)
In-depth Interview	T-PR (Teacher Perceived Role)	Teachers’ perception of their changing role	“I now see myself more as a learning designer and facilitator, rather than just a transmitter of knowledge. It’s a paradigm shift in how we approach teaching”. (Professor W, English Literature)
Document Analysis	C-AI-I (Curriculum-AI Integration)	AI integration in curriculum design	Course syllabus: “20% of the course content will focus on AI-assisted literary criticism practices, blending traditional methods with cutting-edge technology”.
Participatory Observation	TC-AD (Technology Adaptation)	Process of adapting to AI technology	“Observed a clear learning curve effect. Initial weeks showed struggles with AI integration, but by mid-semester, both teachers and students demonstrated markedly improved efficiency”. (Researcher’s reflective journal, Week 10)

Table 2

Potential themes and relevant codes.

No.	Potential Themes	Relevant Codes	Theme Description
1	AI-Enhanced Teaching Practices	AI-LI, AI-LO, AI-IF, T-AI-I, C-AI-I	Reflects how AI directly impacts and changes teaching methods and content delivery
2	Transforming Learning Experiences	S-AI-I, S-ML, AI-GF, S-AI-W, AI-LS, AI-MO	Describes how students interact with AI and how these interactions transform their learning processes and experiences
3	Reconstructing Educational Roles	T-PR, T-AI-L, AI-TD, AI-CP	Discusses the changing roles of teachers and the new skills and professional development required
4	Ethical and Policy Challenges	E-CH, P-CH, AI-ER	Addresses the ethical issues and policy adjustment needs brought about by AI application
5	AI Literacy and Critical Thinking	S-AI-L, AI-LI, AI-CP	Emphasizes the importance of cultivating critical thinking and AI literacy in the AI era
6	Reshaping Educational Spaces and Resources	LR-CH, SP-DI, CL-DY	Reflects the changes in physical and virtual educational spaces and learning resources
7	Innovating Assessment Methods	AS-TR, AI-ER	Explores new assessment methods and standards in the AI era
8	Cultural and Identity Issues	AI-IC, AI-ID	Considers AI’s impact on cultural understanding and learner identity formation
9	New Modes of Collaborative Learning	T-S-AI-I, AI-CL	Describes how AI changes the nature and form of collaborative learning
10	Technology Adaptation and Development	TC-AD, AI-AD, AI-LI	Reflects the process and challenges of educators and learners adapting to AI technology

Appendix A

Table A1

Initial Coding Scheme for Generative AI-Assisted Foreign Language Education.

Data Source	Initial Code	Code Description	Example Quotes
Classroom Observation	AI-LI (AI-assisted Language Input)	AI system providing language input materials	“The AI-generated multi-perspective interpretations of ‘Pride and Prejudice’, including feminist, social class, and marriage views”. (Observation notes, Week 3)
Classroom Observation	AI-LO (AI-assisted Language Output)	AI system assisting student language production	“Students utilized AI writing assistants for essay composition, receiving real-time grammar corrections and stylistic suggestions”. (Observation notes, Week 5)
Classroom Observation	AI-IF (AI Instant Feedback)	Immediate learning feedback provided by AI	“During oral practice, the AI system analyzed students’ pronunciation and intonation in real-time, offering improvement suggestions”. (Observation notes, Week 2)
Classroom Observation	T-AI-I (Teacher-AI Interaction)	Interaction process between teachers and AI	“Professor Z used voice commands to instruct the AI to generate a set of analytical questions on Shakespeare’s sonnets”. (Observation notes, Week 7)
Classroom Observation	S-AI-I (Student-AI Interaction)	Interaction process between students and AI	“Students engaged with the AI chatbot, posing questions about stream of consciousness techniques in modernist literature”. (Observation notes, Week 9)
Classroom Observation	T-S-AI-I (Teacher-Student-AI Interaction)	Three-way interaction among teachers, students, and AI	“Dr. L facilitated a collaborative translation project where students and AI worked together, with ongoing discussions about AI’s translation suggestions”. (Observation notes, Week 11)
In-depth Interview	T-PR (Teacher Perceived Role)	Teachers’ perception of their changing role	“I now see myself more as a learning designer and facilitator, rather than just a transmitter of knowledge. It’s a paradigm shift in how we approach teaching”. (Professor W, English Literature)
In-depth Interview	S-ML (Student Metalearning)	Students’ reflection on their learning process	“AI has made me realize that the key isn’t memorizing information, but learning how to effectively utilize and analyze it. It’s changed how I approach my studies entirely”. (Student L, 3rd year student)
In-depth Interview	AI-AD (AI Advantages)	Perceived advantages of AI systems	“The AI’s ability to provide personalized learning paths is revolutionary. It’s like having a personal tutor for each student, which is impossible in traditional large classes”. (Dr. C, Comparative Literature)
In-depth Interview	AI-LI (AI Limitations)	Perceived limitations of AI systems	“While impressive, the AI sometimes generates content that seems plausible but is factually incorrect. It’s crucial to maintain a critical eye and cross-verify information”. (Student Z, graduate student)
In-depth Interview	E-CH (Ethical Challenges)	Ethical issues arising from AI application	“Ensuring academic integrity in an AI-assisted environment is one of our biggest challenges. We’re still grappling with where to draw the line”. (Professor L, Department Head)
In-depth Interview	P-CH (Policy Challenges)	Policy challenges brought by AI application	“We’re in the process of overhauling our examination and assessment policies to align with AI-era teaching practices. It’s a complex undertaking”. (Dr. W, Academic Affairs)
In-depth Interview	T-AI-L (Teacher AI Literacy)	Teachers’ AI-related competencies	“We’re learning the art of ‘prompting’-how to effectively instruct the AI to get the best educational outcomes. It’s becoming a crucial skill for educators”. (Dr. Z, Western Literary Theory)
In-depth Interview	S-AI-L (Student AI Literacy)	Students’ AI-related competencies	“Beyond language learning, we’re developing critical skills in using and evaluating AI tools. It feels like we’re preparing for a new kind of literacy”. (Student W, 4th year student)
Document Analysis	C-AI-I (Curriculum-AI Integration)	AI integration in curriculum design	Course syllabus: “20% of the course content will focus on AI-assisted literary criticism practices, blending traditional methods with cutting-edge technology”.
Document Analysis	AI-GF (AI-Generated Feedback)	Learning feedback generated by AI systems	AI system log: “Generated 500+ personalized learning recommendations, covering aspects of vocabulary, grammar, and style, tailored to individual student profiles”.
Document Analysis	S-AI-W (Student AI-assisted Work)	Student work completed with AI assistance	Student essay sample: “This paper’s initial structure was developed with AI assistance, followed by substantial human revision and deepening of arguments”.
Document Analysis	AI-ER (AI Evaluation Rules)	Evaluation rules regarding AI use	Departmental policy document: “Students must declare specific portions of assignments completed with AI assistance and explain its contribution to their work”.
Document Analysis	AI-TD (AI Teacher Development)	AI-related teacher development programs	Training document: “All foreign language faculty are required to complete the ‘AI-Assisted Teaching’ workshop series, totaling 40 h of professional development”.
Participatory Observation	CL-DY (Classroom Dynamics)	Classroom interaction dynamics post-AI introduction	“The classroom atmosphere has become more dynamic, with increased student engagement. However, there’s a noticeable tendency towards over-reliance on AI in some discussions”. (Researcher’s reflective journal, Week 8)
Participatory Observation	LR-CH (Learning Resource Changes)	Changes in nature and use of learning resources	“Traditional textbook usage has decreased significantly. There’s a clear shift towards real-time, AI-generated personalized learning materials”. (Researcher’s reflective journal, Week 6)
Participatory Observation	AS-TR (Assessment Transformation)	Transformation of assessment methods	“The assessment strategy has evolved from singular final exams to a more diverse, process-oriented approach, incorporating AI-assisted projects and reflective reports”. (Researcher’s reflective journal, Week 12)
Participatory Observation	TC-AD (Technology Adaptation)	Process of adapting to AI technology	“Observed a clear learning curve effect. Initial weeks showed struggles with AI integration, but by mid-semester, both teachers and students demonstrated markedly improved efficiency”. (Researcher’s reflective journal, Week 10)
Participatory Observation	SP-DI (Spatial Design Impact)	Impact of teaching space design on AI application	“Traditional classroom layouts prove challenging for AI-assisted group collaborations. There’s a growing need for more flexible, technology-integrated space designs”. (Researcher’s reflective journal, Week 4)
Focus Group	AI-CP (AI and Critical Pedagogy)	Discussions on AI’s role in fostering critical thinking	“While AI can provide diverse perspectives, we must ensure it doesn’t replace the development of students’ own critical faculties”. (Professor Y, Focus Group 2)
Focus Group	AI-CL (AI and Collaborative Learning)	Impact of AI on collaborative learning practices	“AI has transformed group projects. Students now collaborate not just with each other, but with AI as a kind of digital team member”. (Dr. L, Focus Group 1)
Focus Group	AI-IC (AI and Intercultural Competence)	AI’s influence on intercultural competence development	“The AI’s ability to provide instant cultural context has both positive and potentially problematic implications for intercultural learning”. (Professor Z, Focus Group 3)
Student Reflective Journals	AI-LS (AI and Learning Strategies)	Changes in learning strategies due to AI integration	“I’ve had to develop new strategies for information synthesis. It’s not just about finding information anymore, but about critically evaluating AI-generated content”. (Student Journal, Week 9)
Student Reflective Journals	AI-MO (AI and Learning Motivation)	AI’s impact on student motivation	“The AI writing assistant has boosted my confidence in expressing complex ideas in English. I feel more motivated to tackle challenging texts”. (Student Journal, Week 7)
Student Reflective Journals	AI-ID (AI and Learner Identity)	Influence of AI on learner identity	“Using AI has made me question what it means to be ‘fluent’ or ‘proficient’ in a language. It’s reshaping how I view my language learning journey”. (Student Journal, Week 11)

References

1. Kasneci, E.; Seßler, K.; Küchemann, S.; Chemann, S.; Bannert, M.; Dementieva, D.; Fischer, F.; Gasser, U.; Groh, G.; Günnemann, S. et al. ChatGPT for good? On opportunities and challenges of large language models for education. Learn. Individ. Differ.; 2023; 103, 102274. [DOI: https://dx.doi.org/10.1016/j.lindif.2023.102274]

2. Wu, T.; He, S.; Liu, J.; Sun, S.; Liu, K.; Han, Q.; Tang, Y. A brief overview of ChatGPT: The history, status quo and potential future development. IEEE CAA J. Autom. Sin.; 2023; 10, pp. 1122-1136. [DOI: https://dx.doi.org/10.1109/JAS.2023.123618]

3. Pan, S.J.; Yang, Q. A survey on transfer learning. IEEE Trans. Knowl. Data Eng.; 2009; 22, pp. 1345-1359. [DOI: https://dx.doi.org/10.1109/TKDE.2009.191]

4. Golonka, E.M.; Bowles, A.R.; Frank, V.M.; Richardson, D.L.; Freynik, S. Technologies for foreign language learning: A review of technology types and their effectiveness. Comput. Assist. Lang. Learn.; 2014; 27, pp. 70-105. [DOI: https://dx.doi.org/10.1080/09588221.2012.700315]

5. Yang, H.; Kyun, S. The current research trend of artificial intelligence in language learning: A systematic empirical literature review from an activity theory perspective. Australas. J. Educ. Technol.; 2022; 38, pp. 180-210. [DOI: https://dx.doi.org/10.14742/ajet.7492]

6. Fryer, L.K.; Nakao, K.; Thompson, A. Chatbot learning partners: Connecting learning experiences, interest and competence. Comput. Hum. Behav.; 2019; 93, pp. 279-289. [DOI: https://dx.doi.org/10.1016/j.chb.2018.12.023]

7. Duan, R.; Kawahara, T.; Dantsuji, M.; Nanjo, H. Cross-lingual transfer learning of non-native acoustic modeling for pronunciation error detection and diagnosis. IEEE ACM Trans. Audio Speech Lang.; 2019; 28, pp. 391-401. [DOI: https://dx.doi.org/10.1109/TASLP.2019.2955858]

8. Ramesh, D.; Sanampudi, S.K. An automated essay scoring systems: A systematic literature review. Artif. Intell. Rev.; 2022; 55, pp. 2495-2527. [DOI: https://dx.doi.org/10.1007/s10462-021-10068-2]

9. Lamb, M. The motivational dimension of language teaching. Lang. Teach.; 2017; 50, pp. 301-346. [DOI: https://dx.doi.org/10.1017/S0261444817000088]

10. Gligorea, I.; Cioca, M.; Oancea, R.; Gorski, A.-T.; Gorski, H.; Tudorache, P. Adaptive Learning Using Artificial Intelligence in e-Learning: A Literature Review. Educ. Sci.; 2023; 13, 1216. [DOI: https://dx.doi.org/10.3390/educsci13121216]

11. Geels, F.W. From sectoral systems of innovation to socio-technical systems: Insights about dynamics and change from sociology and institutional theory. Res. Policy; 2004; 33, pp. 897-920. [DOI: https://dx.doi.org/10.1016/j.respol.2004.01.015]

12. Santana, C.; Albareda, L. Blockchain and the emergence of Decentralized Autonomous Organizations (DAOs): An integrative model and research agenda. Technol. Forecast. Soc. Chang.; 2022; 182, 121806. [DOI: https://dx.doi.org/10.1016/j.techfore.2022.121806]

13. Mishra, P.; Koehler, M.J. Technological pedagogical content knowledge: A framework for teacher knowledge. Teach. Coll. Rec.; 2006; 108, pp. 1017-1054. [DOI: https://dx.doi.org/10.1111/j.1467-9620.2006.00684.x]

14. Mariani, M.M.; Machado, I.; Magrelli, V.; Dwivedi, Y.K. Artificial intelligence in innovation research: A systematic review, conceptual framework, and future research directions. Technovation; 2023; 122, 102623. [DOI: https://dx.doi.org/10.1016/j.technovation.2022.102623]

15. Braun, V.; Clarke, V. Conceptual and design thinking for thematic analysis. Qual. Psychol.; 2022; 9, pp. 3-26. [DOI: https://dx.doi.org/10.1037/qup0000196]

16. Bond, M.; Zawacki-Richter, O.; Nichols, M. Revisiting five decades of educational technology research: A content and authorship analysis of the British Journal of Educational Technology. Br. J. Educ. Technol.; 2019; 50, pp. 12-63. [DOI: https://dx.doi.org/10.1111/bjet.12730]

17. Kulik, J.A.; Fletcher, J.D. Effectiveness of intelligent tutoring systems: A meta-analytic review. Rev. Educ. Res.; 2016; 86, pp. 42-78. [DOI: https://dx.doi.org/10.3102/0034654315581420]

18. Orlikowski, W.J.; Scott, S.V. 10 sociomateriality: Challenging the separation of technology, work and organization. Acad. Manag. Ann.; 2008; 2, pp. 433-474. [DOI: https://dx.doi.org/10.5465/19416520802211644]

19. Weiss, K.; Khoshgoftaar, T.M.; Wang, D. A survey of transfer learning. J. Big Data; 2016; 3, pp. 1-40. [DOI: https://dx.doi.org/10.1186/s40537-016-0043-6]

20. Dessì, D.; Fenu, G.; Marras, M.; Recupero, D.R. Bridging learning analytics and cognitive computing for big data classification in micro-learning video collections. Comput. Hum. Behav.; 2019; 92, pp. 468-477. [DOI: https://dx.doi.org/10.1016/j.chb.2018.03.004]

21. Ouyang, F.; Zheng, L.; Jiao, P. Artificial intelligence in online higher education: A systematic review of empirical research from 2011 to 2020. Educ. Inf. Technol.; 2022; 27, pp. 7893-7925. [DOI: https://dx.doi.org/10.1007/s10639-022-10925-9]

22. Huang, W.; Hew, K.F.; Fryer, L.K. Chatbots for language learning—Are they really useful? A systematic review of chatbot-supported language learning. J. Comput. Assist. Learn.; 2022; 38, pp. 237-257. [DOI: https://dx.doi.org/10.1111/jcal.12610]

23. Zhan, Z.; Fong, P.S.; Mei, H.; Liang, T. Effects of gender grouping on students’ group performance, individual achievements and attitudes in computer-supported collaborative learning. Comput. Hum. Behav.; 2015; 48, pp. 587-596. [DOI: https://dx.doi.org/10.1016/j.chb.2015.02.038]

24. Scherer, R.; Siddiq, F.; Tondeur, J. The technology acceptance model (TAM): A meta-analytic structural equation modeling approach to explaining teachers’ adoption of digital technology in education. Comput. Educ.; 2019; 128, pp. 13-35. [DOI: https://dx.doi.org/10.1016/j.compedu.2018.09.009]

25. Panadero, E. A review of self-regulated learning: Six models and four directions for research. Front. Psychol.; 2017; 8, 422. [DOI: https://dx.doi.org/10.3389/fpsyg.2017.00422]

26. Selwyn, N. What’s the problem with learning analytics?. J. Learn. Anal.; 2019; 6, pp. 11-19. [DOI: https://dx.doi.org/10.18608/jla.2019.63.3]

27. Sarker, S.; Chatterjee, S.; Xiao, X.; Elbanna, A. The sociotechnical axis of cohesion for the IS discipline: Its historical legacy and its continued relevance. MIS Q.; 2019; 43, pp. 695-720. [DOI: https://dx.doi.org/10.25300/MISQ/2019/13747]

28. Lyytinen, K.; Nickerson, J.V.; King, J.L. Metahuman systems= humans+ machines that learn. J. Inf. Technol.; 2021; 36, pp. 427-445. [DOI: https://dx.doi.org/10.1177/0268396220915917]

29. Faraj, S.; Pachidi, S.; Sayegh, K. Working and organizing in the age of the learning algorithm. Inf. Organ.; 2018; 28, pp. 62-70. [DOI: https://dx.doi.org/10.1016/j.infoandorg.2018.02.005]

30. Starkey, L. A review of research exploring teacher preparation for the digital age. Camb. J. Educ.; 2020; 50, pp. 37-56. [DOI: https://dx.doi.org/10.1080/0305764X.2019.1625867]

31. Jain, R.; Garg, N.; Khera, S.N. Effective human—AI work design for collaborative decision-making. Kybernetes; 2023; 52, pp. 5017-5040. [DOI: https://dx.doi.org/10.1108/K-04-2022-0548]

32. Zawacki-Richter, O.; Martin, V.I.; Bond, M.; Gouverneur, F. Systematic review of research on artificial intelligence applications in higher education—where are the educators?. Int. J. Educ. Technol. High. Educ.; 2019; 16, pp. 1-27. [DOI: https://dx.doi.org/10.1186/s41239-019-0171-0]

33. Ouyang, F.; Xu, W.; Cukurova, M. An artificial intelligence-driven learning analytics method to examine the collaborative problem-solving process from the complex adaptive systems perspective. Int. J. Comput.-Support. Collab. Learn.; 2023; 18, pp. 39-66. [DOI: https://dx.doi.org/10.1007/s11412-023-09387-z]

34. Wessel, L.; Baiyere, A.; Ologeanu-Taddei, R.; Cha, J.; Blegind-Jensen, T. Unpacking the difference between digital transformation and IT-enabled organizational transformation. J. Assoc. Inf. Syst.; 2021; 22, pp. 102-129. [DOI: https://dx.doi.org/10.17705/1jais.00655]

35. Johnson, R.B.; Onwuegbuzie, A.J. Mixed methods research: A research paradigm whose time has come. Educ. Res.; 2004; 33, pp. 14-26. [DOI: https://dx.doi.org/10.3102/0013189X033007014]

36. Halverson, L.R.; Graham, C.R.; Spring, K.J.; Drysdale, J.S.; Henrie, C.R. A thematic analysis of the most highly cited scholarship in the first decade of blended learning research. Internet High. Educ.; 2014; 20, pp. 20-34. [DOI: https://dx.doi.org/10.1016/j.iheduc.2013.09.004]

37. Greene, J.C. Is mixed methods social inquiry a distinctive methodology?. J. Mix. Method Res.; 2008; 2, pp. 7-22. [DOI: https://dx.doi.org/10.1177/1558689807309969]

38. Lungu, M. The coding manual for qualitative researchers. Am. J. Qual. Res.; 2022; 6, pp. 232-237. [DOI: https://dx.doi.org/10.29333/ajqr/12085]

39. Viberg, O.; Hatakka, M.; Bälter, O.; Mavroudi, A. The current landscape of learning analytics in higher education. Comput. Hum. Behav.; 2018; 89, pp. 98-110. [DOI: https://dx.doi.org/10.1016/j.chb.2018.07.027]

40. Gašević, D.; Dawson, S.; Siemens, G. Let’s not forget: Learning analytics are about learning. TechTrends; 2015; 59, pp. 64-71. [DOI: https://dx.doi.org/10.1007/s11528-014-0822-x]

41. Sghir, N.; Adadi, A.; Lahmer, M. Recent advances in Predictive Learning Analytics: A decade systematic review (2012–2022). Educ. Inf. Technol.; 2023; 28, pp. 8299-8333. [DOI: https://dx.doi.org/10.1007/s10639-022-11536-0] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36571084]

42. Dedrick, R.F.; Ferron, J.M.; Hess, M.R.; Hogarty, K.Y.; Kromrey, J.D.; Lang, T.R.; Niles, J.D.; Lee, R.S. Multilevel modeling: A review of methodological issues and applications. Rev. Educ. Res.; 2009; 79, pp. 69-102. [DOI: https://dx.doi.org/10.3102/0034654308325581]

43. Grant, C.; Osanloo, A. Understanding, selecting, and integrating a theoretical framework in dissertation research: Creating the blueprint for your “house”. Adm. Issues J.; 2014; 4, 4. [DOI: https://dx.doi.org/10.5929/2014.4.2.9]

44. Celik, I.; Dindar, M.; Muukkonen, H.; Järvelä, S. The promises and challenges of artificial intelligence for teachers: A systematic review of research. TechTrends; 2022; 66, pp. 616-630. [DOI: https://dx.doi.org/10.1007/s11528-022-00715-y]

45. Collins, K.M.; Onwuegbuzie, A.J.; Jiao, Q.G. Prevalence of mixed-methods sampling designs in social science research. Eval. Res. Educ.; 2006; 19, pp. 83-101. [DOI: https://dx.doi.org/10.2167/eri421.0]

46. Richards, K.A.R.; Hemphill, M.A. A practical guide to collaborative qualitative data analysis. J. Teach. Phys. Educ.; 2018; 37, pp. 225-231. [DOI: https://dx.doi.org/10.1123/jtpe.2017-0084]

47. Broccardo, L.; Zicari, A.A.N.; Jabeen, F.; Bhatti, Z.A. How digitalization supports a sustainable business model: A literature review. Technol. Forecast. Soc. Chang.; 2023; 187, 122146. [DOI: https://dx.doi.org/10.1016/j.techfore.2022.122146]

48. Ivankova, N.V.; Creswell, J.W.; Stick, S.L. Using mixed-methods sequential explanatory design: From theory to practice. Field Method; 2006; 18, pp. 3-20. [DOI: https://dx.doi.org/10.1177/1525822X05282260]

49. Gerring, J. What is a case study and what is it good for?. Am. Political Sci. Rev.; 2004; 98, pp. 341-354. [DOI: https://dx.doi.org/10.1017/S0003055404001182]

50. Yin, R.K. Case Study Research: Design and Methods; Sage Publications: London, UK, 2009; pp. 16-20.

51. Adeoye-Olatunde, O.A.; Olenik, N.L. Research and scholarly methods: Semi-structured interviews. J. Am. Coll. Clin. Pharm.; 2021; 4, pp. 1358-1367. [DOI: https://dx.doi.org/10.1002/jac5.1441]

52. Lewis, S. Qualitative inquiry and research design: Choosing among five approaches. Health Promot. Pract.; 2015; 16, pp. 473-475. [DOI: https://dx.doi.org/10.1177/1524839915580941]

53. Hays, D.G.; McKibben, W.B. Promoting rigorous research: Generalizability and qualitative research. J. Couns. Dev.; 2021; 99, pp. 178-188. [DOI: https://dx.doi.org/10.1002/jcad.12365]

54. Zaman, R.; Jain, T.; Samara, G.; Jamali, D. Corporate governance meets corporate social responsibility: Mapping the interface. Bus. Soc.; 2022; 61, pp. 690-752. [DOI: https://dx.doi.org/10.1177/0007650320973415]

55. Roberts, R.E. Qualitative Interview Questions: Guidance for Novice Researchers. Qual. Rep.; 2020; 25, pp. 3185-3203. [DOI: https://dx.doi.org/10.46743/2160-3715/2020.4640]

56. Kim, H.; Sefcik, J.S.; Bradway, C. Characteristics of qualitative descriptive studies: A systematic review. Res. Nurs. Health; 2017; 40, pp. 23-42. [DOI: https://dx.doi.org/10.1002/nur.21768]

57. Korstjens, I.; Moser, A. Series: Practical guidance to qualitative research. Part 4: Trustworthiness and publishing. Eur. J. Gen. Pract.; 2018; 24, pp. 120-124. [DOI: https://dx.doi.org/10.1080/13814788.2017.1375092]

58. Braun, V.; Clarke, V. Using thematic analysis in psychology. Qual. Res. Psychol.; 2006; 3, pp. 77-101. [DOI: https://dx.doi.org/10.1191/1478088706qp063oa]

59. Nowell, L.S.; Norris, J.M.; White, D.E.; Moules, N.J. Thematic analysis: Striving to meet the trustworthiness criteria. Int. J. Qual. Methods; 2017; 16, 1067544023. [DOI: https://dx.doi.org/10.1177/1609406917733847]

60. Kiger, M.E.; Varpio, L. Thematic analysis of qualitative data: AMEE Guide No. 131. Med. Teach.; 2020; 42, pp. 846-854. [DOI: https://dx.doi.org/10.1080/0142159X.2020.1755030] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/32356468]

61. De Casterlé, B.D.; Gastmans, C.; Bryon, E.; Denier, Y. QUAGOL: A guide for qualitative data analysis. Int. J. Nurs. Stud.; 2012; 49, pp. 360-371. [DOI: https://dx.doi.org/10.1016/j.ijnurstu.2011.09.012]

62. Alam, M.K. A systematic qualitative case study: Questions, data collection, NVivo analysis and saturation. Qual. Res. Organ. Manag.; 2021; 16, pp. 1-31. [DOI: https://dx.doi.org/10.1108/QROM-09-2019-1825]

63. Holton, J.A. Grounded theory as a general research methodology. Grounded Theory Rev.; 2008; 7, pp. 67-93. Available online: https://groundedtheoryreview.com/2008/06/30/grounded-theory-as-a-general-research-methodology/ (accessed on 4 September 2024).

64. Naeem, M.; Ozuem, W.; Howell, K.; Ranfagni, S. A step-by-step process of thematic analysis to develop a conceptual model in qualitative research. Int. J. Qual. Methods; 2023; 22, 2139372957. [DOI: https://dx.doi.org/10.1177/16094069231205789]

65. Braun, V.; Clarke, V. Toward good practice in thematic analysis: Avoiding common problems and be (com) ing a knowing researcher. Int. J. Transgender Health; 2023; 24, pp. 1-6. [DOI: https://dx.doi.org/10.1080/26895269.2022.2129597] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36713144]

66. Tracy, S.J. Qualitative quality: Eight “big-tent” criteria for excellent qualitative research. Qual. Inq.; 2010; 16, pp. 837-851. [DOI: https://dx.doi.org/10.1177/1077800410383121]

67. Bethencourt-Aguilar, A.; Castellanos-Nieves, D.; Sosa-Alonso, J.E.; Area-Moreira, M. Use of generative adversarial networks (GANs) in educational technology research. J. New Approaches Educ. Res.; 2023; 12, pp. 153-170. [DOI: https://dx.doi.org/10.7821/naer.2023.1.1231]

68. Siau, K.; Wang, W. Artificial intelligence (AI) ethics: Ethics of AI and ethical AI. J. Database Manag.; 2020; 31, pp. 74-87. [DOI: https://dx.doi.org/10.4018/JDM.2020040105]

69. Tsai, Y.S.; Poquet, O.; Gašević, D.; Dawson, S.; Pardo, A. Complexity leadership in learning analytics: Drivers, challenges and opportunities. Br. J. Educ. Technol.; 2019; 50, pp. 2839-2854. [DOI: https://dx.doi.org/10.1111/bjet.12846]

70. Rasul, T.; Nair, S.; Kalendra, D.; Robin, M.; de Oliveira Santini, F.; Ladeira, W.J.; Sun, M.; Day, I.; Rather, R.A.; Heathcote, L. The role of ChatGPT in higher education: Benefits, challenges, and future research directions. J. Appl. Learn. Teach.; 2023; 6, pp. 41-56. [DOI: https://dx.doi.org/10.37074/jalt.2023.6.1.29]

71. Dimitrieska, V. Language teacher identity construction: Reflective conversation. J. Lang. Identity Educ.; 2022; 23, pp. 771-786. [DOI: https://dx.doi.org/10.1080/15348458.2022.2054418]

72. Andersen, R.; Mørch, A.I.; Litherland, K.T. Collaborative learning with block-based programming: Investigating human-centered artificial intelligence in education. Behav. Inf. Technol.; 2022; 41, pp. 1830-1847. [DOI: https://dx.doi.org/10.1080/0144929X.2022.2083981]

73. Palvia, S.; Aeron, P.; Gupta, P.; Mahapatra, D.; Parida, R.; Rosner, R.; Sindhi, S. Online education: Worldwide status, challenges, trends, and implications. J. Glob. Inf. Technol. Manag.; 2018; 21, pp. 233-241. [DOI: https://dx.doi.org/10.1080/1097198X.2018.1542262]

74. Ertmer, P.A.; Ottenbreit-Leftwich, A.T. Teacher technology change: How knowledge, confidence, beliefs, and culture intersect. J. Res. Technol. Educ.; 2010; 42, pp. 255-284. [DOI: https://dx.doi.org/10.1080/15391523.2010.10782551]

75. Lai, C.; Gu, M. Self-regulated out-of-class language learning with technology. Comput. Assist. Lang. Learn.; 2011; 24, pp. 317-335. [DOI: https://dx.doi.org/10.1080/09588221.2011.568417]