The impact of multimodal educational content created using generative artificial intelligence on the quality of the educational process: a PLS-SEM analysis
Main Article Content
Abstract
This article examines the impact of multimodal educational content created using generative artificial intelligence (AI-generated video lectures, synthesized speech, and automatic subtitles) on the quality of the educational process. The relevance of this research stems from the lack of quantitatively verified models that link the structure of automatically generated content to pedagogical outcomes via intermediate mechanisms – social interaction, instructor support, and language localization. The aim is to develop and preliminarily test a PLS-SEM model for assessing this impact. A production pipeline for video lectures has been developed, incorporating a mandatory review stage by the academic supervisor, and a conceptual model of structural modeling using the partial least squares method has been constructed, encompassing three groups of independent content variables, three mediators, three dependent variables of learning quality, and three contextual moderators. The model was preliminarily tested on a simulated sample using SmartPLS, with bootstrap significance testing. Acceptable convergent and discriminant validity were confirmed for reflective constructs; the collinearity of formative indicators does not exceed the permissible threshold. The influence of visual media content on the formation of a socio-interactive environment and the marginal influence of audio content on language localization have been confirmed, whilst the influence of mediators on learning outcomes proved to be statistically insignificant, suggesting insufficient sample power rather than the absence of a real relationship in the general population. The results support the feasibility of a hybrid, human-centered architecture for the automated production of educational videos and outline a direction for further research with an expanded sample of respondents for confirmatory analysis.

