Digital Translanguaging and Cognitive Ease: An Observational Analysis of Postgraduate Students’ Engagement with Multimodal AI for Simplified Academic Conceptualization

HRMARS - The emergence of Generative Artificial Intelligence (GenAI) has introduced new linguistic dynamics in higher education, particularly within multilingual academic settings. While highly sophisticated models such as ChatGPT and DeepSeek dominate the discourse on academic AI, postgraduate students are increasingly exhibiting a preference for "plain language" or "layman-term" tools like Doubao and Dola. This study investigates the strategic use of AI translanguaging tools among postgraduate students in Malaysia through a qualitative approach involving naturalistic observation and semi-structured interviews. Grounded in Cognitive Load Theory and the AI-integrated Social Media Adoption (AISMA) Model (Emon, 2026), the research explores how the linguistic register of AI serves as a cognitive scaffold. Findings reveal that students utilize Doubao and Dola for initial conceptual sense-making to lower their Affective Filter, while reserving ChatGPT and DeepSeek for stylistic polishing and logical validation. Furthermore, the study identifies Platform Embeddedness—specifically within the WeChat/Weixin ecosystem—as a primary driver for tool selection, facilitating "in-situ" translanguaging that reduces platform friction. The results suggest that for postgraduates, the utility of AI is defined not by its formal academic rigor, but by its ability to bridge the gap between everyday linguistic repertoires and specialized scholarly discourse. This paper concludes by proposing a hybrid workflow model that integrates pedagogical translanguaging with AI-assisted research practices.