Nonlinear Dynamic Language Learning Theory in AI-Mediated EFL: From Theory to Practice

Grounded in a critical-realist ontology and a pragmatic-constructivist epistemology, this study operationalizes Nonlinear Dynamic Language Learning Theory (NDLLT) in AI-mediated EFL classrooms and empirically examines motivation as a fluctuating, history-dependent system. A 12-week randomized controlled trial (N = 784; CEFR B2–C1) compared three collaborative AI conditions (AI-enhanced Socrative, team-based Kahoot!, adaptive Duolingo + collaborative production) with an active CALL control. Outcomes included TOEFL iBT skills, a 50-item NDLLS motivation scale, an 18-item feedback survey, and interviews. MANCOVA/ANCOVA tested group differences; cross-lagged structural models estimated coupling between proficiency gains and motivational change; nonlinear time-series analyses (e.g., recurrence quantification, detrended fluctuation analysis) characterized attractor strength, variability, and phase shifts. Relative to CALL, AI conditions produced larger gains in reading and writing and more time in high-engagement attractor states, moderated by emotion regulation and peer collaboration. Engagement micro-variability prospectively predicted proficiency gains, consistent with NDLLT’s phase-shift hypothesis. Implementation fidelity (≥90%) and accessibility/fairness safeguards supported validity. Findings depict proficiency and motivation as co-evolving trajectories within learner–AI–peer ecologies and argue for proficiency-sensitive scaffolding that tunes control parameters (challenge–skill balance, feedback timing, peer coupling) rather than prescribing linear sequences. The study offers design and evaluation principles for equitable, scalable AI integration in EFL contexts.