Ai-Driven Personalization of Power System Learning Modules Using Student Personas based on Behavioral Analysis of Grid Performance
Main Article Content
Abstract
The current advancement of artificial intelligence in educational applications has led to the emergence of previously unexplored possibilities in personalised learning experiences, specifically due to Large Language Models (LLMs). Most uses of LLMs in education have not been rigorously tested with statistical models to establish the extent to which they have been effective in making learners more adaptable. This work examines the use of statistical models of student personas in integration with LLMs to produce customised educational content responding to learner profiles. We applied a mixed methods quasi-experimental research design in which we used means clustering to group students into separate clusters in relation to cognitive style, motivation and past performance, and then we validated our groups by validating the clusters using silhouette analysis and ANOVA. The content created by LLM was then tailored to the personality traits of each of the personas and contrasted with generic instructional materials in both treatment and control groups. Learning adaptability was a multi-item scale measure and was analysed in both multiple regression models and interaction effects of variables. The findings have indicated that personas in LLM content offered to the students depicted a vastly high score of adaptability and interest rate to students compared to their counterparts in the control cell. The result of the regression analysis proved that the type of persona and the type of interaction with personalised content showed good strength in predicting adaptability (p < .01). These results indicate that educationally informed learner modelling, combined with powerful generative AI, has a potential to boost responsiveness to a large extent. The research work is important to the discipline in the sense that it provides empirical evidence of the importance of integrating behavioural statistics with new AI tools in learning. The implications to educational technologists, learning designers and behavioural statisticians are mentioned, and the recommendations to scale ethically, such innovations are proposed