Profiles and Transitions of Psychological Adaptation in AI-Assisted Japanese Language Learning

Synthesis: Yuanyuan Wu (2026) examined how learners psychologically adapt to AI-assisted Japanese language learning, using a three-wave longitudinal survey of 457 learners across one academic semester with technostress and psychological resilience as indicators in latent profile and latent profile transition analyses. Three profiles were consistently identified — maladaptive adaptation, moderate adaptation, and positive adaptation — with the maladaptive group shrinking from 26.48% to 17.74% and the positive group growing from 27.35% to 35.73% across the semester. The positive-adaptation profile showed the greatest temporal stability, while maladaptive learners most commonly transitioned to moderate adaptation. At Time 3, positively adapting learners reported the highest Japanese learning Self-Efficacy and the lowest learning burnout, with the reverse pattern for maladaptive learners, demonstrating that psychological adaptation to AI-assisted language learning is heterogeneous, dynamic, and meaningfully tied to confidence and burnout.

Key Findings

• Three stable psychological-adaptation profiles: Learners were consistently classifiable into maladaptive, moderate, and positive adaptation profiles across all three waves, indicating a conceptually stable underlying pattern of adaptation to AI-assisted study.
• Adaptation is dynamic, not fixed: The maladaptive profile decreased from 26.48% to 17.74% while the positive profile increased from 27.35% to 35.73% over the semester; positive-adaptation learners were most stable, and maladaptive learners most commonly shifted to moderate adaptation.
• Profiles predict confidence and burnout: At Time 3, positively adapting learners reported the highest Japanese learning self-efficacy and lowest burnout; maladaptive learners showed the reverse — profiles are practically meaningful for learning outcomes.
• Technostress declined and resilience rose: All five technostress dimensions declined from Time 1 to Time 3 while psychological resilience rose modestly (2.03 to 2.19), consistent with a gradual easing of technological strain as learners gained experience.
• Sample context: 457 learners at Time 1 (76.1% female), mostly enrolled in non-major Japanese courses (56.0%) with intermediate proficiency (45.7%); 62.1% had no prior AI training, and the most common AI-use frequency was 3–4 times per week.

What this means for practice

• Learners. Track your confidence and workload rather than your tool usage, because at Time 3 the positively adapting profile showed the highest Japanese learning Self-Efficacy and the lowest burnout, while the maladaptive profile showed the reverse.
• Learners. Treat an early bad patch as a stage rather than a verdict: learners in the maladaptive profile most often moved to moderate adaptation rather than jumping straight to positive adaptation, and adaptation is heterogeneous and dynamic across a semester.
• Instructors. Screen for learners whose technostress outweighs their psychological resilience instead of assuming AI tools benefit everyone, since 17.74% of learners remained maladaptive at Time 3 and reported the least favorable motivational and emotional outcomes.
• Instructors. Strengthen adaptive resources rather than adding more tools — fostering confidence and regulating self-regulated AI use is what shifted learners toward positive adaptation under a demand–resource balance.
• Designers. Design against avoidable overload: with 62.1% of learners reporting no prior AI training and 3–4 times per week the most common use frequency, information density and dependence management belong in the learning design, not only in onboarding.

Limitations

• All focal measures are self-reported and the authors flag common method variance and subjective response tendencies; no behavioral indicators of AI use or teacher evaluations of engagement were collected.
• The sample is 457 learners of Japanese at public universities in mainland China (76.1% female; 56.0% in non-major courses; 45.7% intermediate), so the profile structure's generalizability to other languages, institutions, and cultural settings remains to be established.
• Coverage is a single academic semester (three waves roughly 8 weeks apart); only 366 learners completed all three waves (80.09% retention), and whether adaptive improvements persist beyond one semester is unknown.
• The mechanisms the authors invoke — AI literacy, self-regulated learning, routine integration, and perceived teacher support — were not directly measured, so profile membership is not shown to cause the self-efficacy and burnout differences, and other explanatory evidence cited comes from EFL settings that differ in target language and indicators.

Citation

Wu, Y. (2026). Profiles and transitions of psychological adaptation in AI-assisted Japanese language learning. Frontiers in Psychology, 17, 1837484.