NAVIGATING THE GENERATIVE AI PARADOX IN MULTIMODAL COMPOSING: INDONESIAN EFL STUDENTS' PERCEPTIONS AND SELF-REGULATORY ANXIETIES

Abstract

This study investigates Indonesian undergraduate English as a Foreign Language (EFL) students’ perceptions of integrating generative artificial intelligence (GAI) tools into digital multimodal text-writing practices. Employing an exploratory, descriptive quantitative design, survey data were collected from 101 undergraduate students at a private university in Indonesia using a structured 5-point Likert scale questionnaire framed around the Technology Acceptance Model (TAM) and Self-Regulated Learning (SRL) theories. The descriptive statistical results reveal a distinct pedagogical paradox. Students expressed high technology acceptance regarding GAI’s technical affordances, accessibility, and time-saving capabilities (M=3.96), noting that these tools significantly optimize linguistic precision, professional layout, and semiotic resource orchestration (M=3.58). However, the subscale analysis also reveals profound self-regulatory and ethical anxieties. Students demonstrated a critical, defensive awareness regarding the risks of voice erasure due to over-reliance (M = 4.21, systemic informational inaccuracies (M = 3.98), and long-term writing skill attrition (M = 4.05). These findings indicate that while GAI serves as a powerful prompt-driven mediator, high technology acceptance does not inherently result in successful pedagogical integration. The study concludes that higher education institutions must design process-oriented curricula that explicitly couple GAI adoption with explicit AI literacy frameworks and strategic teacher guidance to ensure students act as critical editors rather than passive consumers of automated content.