Rethinking Student Engagement in the Age of AI-Mediated Interaction
IEEE Revista Iberoamericana de Tecnologias del Aprendizaje · 2026 · 21 · 601-615
DOI: 10.1109/rita.2026.3746528 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Revista Iberoamericana de Tecnologias del Aprendizaje, 2026. Verfasst von Manuel Chaves-Maza. Die Arbeit nennt 103 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Universidad Pablo de Olavide/Consorcio de Bibliotecas Universitarias de Andalucía.
Abstract (Originalsprache)
Conventional course delivery rarely adapts to the individual learner, and quantitative service courses such as Statistics combine high failure rates with frequent absence from the examination: in the cohort studied here only 76.6% of 273 enrolled students sat the first sitting and almost one in three of those failed. What is not known is how students actually use AI learning support placed inside such a course. This paper reports a zero-cost field deployment of topic-aligned interactive exercises plus a retrieval-augmented conversational tutor (an open large language model grounded in the course materials, rendered as a 3D avatar) in two undergraduate Business Statistics courses. The research method is a multi-source quantitative observational learning-analytics design. Authentic, unsupervised platform logs, cleaned with a protocol fixed before analysis, were triangulated with the cohort’s institutional grades (273 enrolled; 70.8% pass; mean 5.33/10) and a pilot satisfaction survey; the logs yielded 310 exercise completions from 82 students, of which 153 from 53 students form the primary analytical set. Non-parametric tests (Spearman, Kruskal-Wallis, Wilcoxon) with student-level bootstrap confidence intervals and a Poisson model with quiz fixed effects provide proximal interaction evidence rather than a causal grade claim. Engagement was voluntary and recurrent: 69.8% of Statistics users returned for two or more attempts and 58.5% came back in a different calendar week. The exercises revealed a robust internal structure of curricular difficulty (hardest topic 4.29/10, 95% CI [3.27, 5.37]; Kruskal-Wallis p =.002). Students with a lower first score practised more (Spearman rho = -0.325, 95% CI [-0.566, -0.058]), but that association did not survive normalisation within the same quiz, indicating that practice followed task difficulty rather than relative underperformance. Repeating the same quiz produced no reliably detectable gain. P-values are nominal and unadjusted.
Förderung
- Universidad Pablo de Olavide/Consorcio de Bibliotecas Universitarias de Andalucía
Zitieren
APA
Chaves-Maza, M. (2026). Rethinking Student Engagement in the Age of AI-Mediated Interaction. IEEE Revista Iberoamericana de Tecnologias del Aprendizaje, 21, 601-615. https://doi.org/10.1109/rita.2026.3746528
Vancouver
Chaves-Maza M. Rethinking Student Engagement in the Age of AI-Mediated Interaction. IEEE Revista Iberoamericana de Tecnologias del Aprendizaje. 2026;21:601-615. doi:10.1109/rita.2026.3746528
BibTeX
@article{ChavesMaza2026,
author = {Chaves-Maza, Manuel},
title = {Rethinking Student Engagement in the Age of AI-Mediated Interaction},
journal = {IEEE Revista Iberoamericana de Tecnologias del Aprendizaje},
year = {2026},
volume = {21},
pages = {601--615},
doi = {10.1109/rita.2026.3746528}
}
EUVIMED