Profiling Beliefs about Multiliteracies in AI-Saturated Language Education

Abstract

As generative AI becomes embedded in everyday language learning, the conceptual adequacy of multiliteracies frameworks remains underexamined. This study profiles how students and instructors reinterpret multiliteracies within an AI-saturated English language program at a single comprehensive university. Drawing on a convergent mixed-method design, data were collected over one semester from 178 undergraduate EFL students and 9 instructors. Instruments included a newly developed Multiliteracies-in-AI Scale (MIAS), stimulated-recall interviews, and a corpus of pre- and post-intervention essays. Latent profile analysis identified three belief configurations: (1) Instrumentalist Integrators (AI as efficiency tool), (2) Critical Orchestrators (AI as multimodal co-designer requiring oversight), and (3) Displaced Traditionalists (AI as threat to textual authenticity). Regression modeling revealed that belief profiles significantly predicted observable AI use patterns, including prompt complexity, revision depth, and multimodal integration (p < .01). Notably, students classified as Critical Orchestrators demonstrated higher intertextual diversity and more frequent cross-modal redesign (text–image–data integration) in final submissions. A short, four-week algorithmic literacy module shifted 27% of Instrumentalist Integrators toward the Critical Orchestrator profile. The findings challenge static interpretations of multiliteracies by demonstrating that AI saturation produces differentiated epistemic orientations that are measurable, behaviorally traceable, and pedagogically malleable. We propose a reconceptualization of multiliteracies as dynamic belief–practice constellations shaped by algorithmic co-authorship rather than solely by multimodal textuality.