AI Adoption in Language Education: Learner–Instructor Differences and the Affordance–Accountability Asymmetry

This study examines whether artificial intelligence (AI) adoption operates similarly across learner and instructor roles in language education. Grounded in Rogers’ Diffusion of Innovations Theory, it used an explanatory sequential mixed-methods design. The quantitative phase included 415 participants—276 international learners of Turkish and 139 instructors—while the qualitative phase comprised eight focus groups with 22 learners and 22 instructors. Multi-group structural equation modeling, measurement invariance, ordinal logistic regression, and the Framework Method were used. Relative advantage, compatibility, trialability, and observability showed positive bivariate associations with attitudes toward AI in both groups, although predictor overlap limited interpretation of unique simultaneous effects. No significant between-group differences were detected in the structural paths. Role differences instead emerged in construct organization: complexity components were closely related among learners but more differentiated among instructors, whereas attitude and intention were distinguishable among learners but largely overlapped among instructors. Although both attitude and intention were positively associated with self-reported AI use, intention added little explanatory value beyond attitude when modeled jointly. Qualitative findings showed that similar AI affordances were evaluated through responsibility frameworks. Conceptualized as affordance–accountability asymmetry, this pattern suggests that responsible and sustainable AI integration requires alignment between technological opportunities and role-specific pedagogical responsibilities.