How do personalization methods used in learn-by-mistakes systems affect learning gains?

Abstract

Özyurt and al. (2013) identify that for intelligent learning management systems (iLMS) to be effective, further learning must occur after students’ mistakes. Personalization stays effective when students get accurate results that will keep their motivation high. UZWEBMAT’s (Turkish abbreviation of Adaptive and INtelligent WEB-based MAThematics teaching-learning system) evaluation yields similar results that students maintain high motivations when making mistakes. Differently, Aboderin and al. (2015) tested elearning in Nigeria and found students still learning with low motivation, indicating that students need technical help from teachers (50% of students having trouble logging in and out), but once familiar with the system learning improved. Kochmar and al. (2020) evaluate Korbit, an intelligent tutoring system that offers personalized learning to learn new skills, and conclude that deep and shallow personalization (specific and generic articles from Wikipedia respectively) both result in higher learning gains than a baseline model. More personalization yielded more learning gains for all groups and considerably helped low-ability students. Similarly, Myneni and al. (2013) detail low-ability students receiving more help than other students with especially high learning gains. Low-ability students that are demotivated initially but increase their motivation once exposed to the Rebolledo-Mendez and al. (2006)’s M-Ecolab typically experience the most learning due to their ability to seek challenges, which explains the increase of motivation throughout the interaction. Based on these studies we observed that students with high motivation before using an iLMS usually do not seek challenges, resulting in low learning gains. Overall, iLMS have more effective personalization strategies for low-ability students.