Authors:Zaynab B. Raji-Ellams and Panu Abosede Sewedo mniti
Department of English and Literary Studies, Bayero University Kano and Institute of Translation Studies, University of Ilorin
Date: 01/11/2025
The mediating role of an Artificial Intelligence (AI) tutor in early language acquisition was investigated using a quantitative pilot study. A 100-turn conversational corpus was generated through a simulated interaction between two AI instances: a "tutor" programmed to provide syntactic recasting and a "learner" programmed to produce predictable overregularization errors (e.g., "goed" for "went"). Under baseline conditions without intervention, the simulated learner exhibited a high morphological Error Rate of 90%. Following a sustained 60-turn intervention phase where the tutor provided consistent feedback, the learner's Error Rate decreased to 30% in the final phase of the study. Furthermore, the learner's Correction Uptake Rate i.e., the use of a correct form following a recast, rose to 71% post-intervention. The interaction was analyzed using a three-phase mediational framework (Collection, Analysis, Action), and it was found that the AI's consistent, data driven feedback loop was directly correlated with the positive change in the learner's performance, demonstrating a computationally sound model for personalized linguistic scaffolding.
Download PDF