Exploring The Effect of Self-Efficacy on Technological Acceptance of Artificial Intelligence Usage in L2 Education: A SEM-PLS Analysis

Abstract

Despite the rapid integration of Artificial Intelligence (AI) into language education, the psychological factors driving AI acceptance among secondary-level learners remain poorly understood, particularly in developing-country contexts where ethical AI governance frameworks are underdeveloped. This study examined the effect of Self-Efficacy (SE) on Perceived Ease of Use (PE) and Perceived Usefulness (PU) of AI tools within the Technology Acceptance Model (TAM) framework, among junior and senior high school foreign language learners in Indonesia. Using a quantitative, non-experimental survey design, data were collected from 63 secondary school students who had direct prior experience using AI-based language tools. The instrument adapted Davis’ (1986) TAM questionnaire and Bandura’s (1997) self-efficacy scale and was analysed via Partial Least Squares Structural Equation Modelling (PLS-SEM) using SmartPLS 4.0. Measurement model results confirmed satisfactory reliability (CR ≥ 0.899; Cronbach’s α ≥ 0.826) and validity (AVE ≥ 0.748; HTMT < 0.90) across all constructs. Structural model analysis revealed that SE significantly and positively predicted both PE (β = 0.660, large effect: f² = 0.772) and PU (β = 0.591, large effect: f² = 0.536), with the model explaining 43.6% and 34.9% of the variance in PE and PU, respectively. These findings extend TAM by establishing self-efficacy as a robust psychological antecedent of AI technology acceptance and constitute the first study to integrate ethical AI use, a self-efficacy predictor model, and the secondary foreign language education context in Indonesia. Implications for curriculum design, teacher training, and inclusive AI policy are discussed.