```html

1 Introduction

The development of Generative Artificial Intelligence (GAI) has profoundly impacted the educational ecosystem, particularly in the field of foreign language education. GAI technologies such as DeepSeek and ChatGPT, through features like instant interactive feedback, multilingual recognition, and multimodal generation, have transformed how students engage with knowledge, making teaching more personalized, immediate, and contextualized. In multilingual and multicultural settings, foreign language education must challenge traditional monolingualism, respect learners' individual needs, and encourage them to mobilize linguistic and cognitive resources to construct knowledge and cultural literacy systems. The integration of GAI helps expand translanguaging practices for teachers and students, stimulating learner interest and enhancing their agency. Although research on GAI applications in foreign language education is growing, discussions on learners' translanguaging practices and agency remain insufficient. This study focuses on English Medium Instruction (EMI) classrooms in China, exploring how students engage in translanguaging practices with GAI Chatbots and analyzing their agency characteristics.

2 Literature Review

This section examines the application of translanguaging practices in EMI classrooms and their impact on learner agency. Translanguaging, as a critical theory, emphasizes the dynamic and multimodal nature of language, allowing learners to integrate diverse resources to construct meaning. In EMI classrooms, translanguaging facilitates students' natural use and development of the target language while acquiring subject content. Learner agency refers to individuals' capacity to autonomously regulate their actions to achieve goals within specific sociocultural contexts, characterized by sociocultural mediation, situatedness, complexity, and dynamism. Advances in GAI technology provide new expressive spaces for translanguaging, enabling learners in EMI classrooms to autonomously engage in translanguaging practices and mobilize various resources to acquire disciplinary knowledge. This study aims to explore the forms of translanguaging practices and agency characteristics exhibited by learners during interactions with GAI Chatbots in EMI classrooms, offering insights for educators to effectively integrate translanguaging practices in GAI-enhanced EMI settings and foster learner agency.

3 Research Design

This section outlines the research design, including research questions, participants, data collection, and analytical methods. The study focuses on learners' translanguaging practices and agency manifestations during interactions with GAI Chatbots. Participants include 50 undergraduate students engaged in classroom activities facilitated by GAI Chatbots. Data collection comprises human-AI interaction logs, reflective journals, and open-ended feedback centered on themes of "Language and Culture" and "English Learning." Data analysis employs thematic analysis, with coding applied to translanguaging practices and learner agency.

4 Findings and Discussion

This chapter explores learners' translanguaging practices and agency in GAI-enhanced EMI classrooms. Findings reveal that students exhibit rich multilingual and multimodal translanguaging practices, including language compensation, term migration, emphasis highlighting, and cognitive simplification, as well as multimodal practices such as emojis, images, video links, and music. These practices facilitate integrated learning of language and subject content, improving both language proficiency and disciplinary knowledge acquisition. GAI Chatbots provide a unique translanguaging space that sparks students' imagination and creativity while fostering critical thinking.

In GAI-enabled learning environments, students demonstrate four dimensions of agency: 1) Autonomous planning and regulation—students tailor learning goals, content, and pace based on personal needs and preferences; 2) Critical evaluation and feedback—students actively assess information from GAI Chatbots and propose improvements; 3) Resource integration and utilization—students proactively explore and leverage GAI-mediated resources to meet learning and communicative needs; 4) Situational awareness and adaptation—students exhibit acute sensitivity and adaptability to sociocultural contexts. These agency traits position learners as self-directed, reflective, resourceful, and adaptable agents rather than passive recipients of information. Fostering such agency is crucial for enhancing educational quality.

5 Conclusion

This chapter summarizes findings on integrating GAI Chatbots in EMI classrooms. Results indicate that GAI Chatbots create an inclusive translanguaging space that promotes synergistic development of linguistic and disciplinary competencies through human-AI interaction, alongside emergent multilingual phenomena. Learners demonstrate high agency through multimodal meaning-making. The study offers pedagogical implications, including leveraging GAI's translanguaging potential to reconcile monolingual policies with practical needs, activating learner agency, and documenting student-AI interactions. Educators are cautioned about GAI's potential drawbacks, emphasizing the importance of originality and imagination. Limitations include unmeasured participant language proficiency and reliance on qualitative data from four activities, suggesting future longitudinal studies with language proficiency assessments.

``` (Note: The translation preserves the original HTML tags and structure while converting the Chinese text to fluent, academic English. Key terms like "translanguaging practices" and "learner agency" are maintained as standard concepts in applied linguistics research.)