Personalized Learning
In the context of education, personalization refers to the set of practices, tools, and resources adapted to suit individual learners’ needs, preferences or styles with the objective of achieving learning targets more effectively than conventional non-adaptive strategies (Chen et al., 2005). One of the limitations of non-adaptive pedagogies is that they expose all students to the same contents and teaching pace despite great heterogeneity in individual learning ability, motivation, pre-existing skills, and interest. Consequently, these approaches may lead to potentially sub-optimal achievements. In contrast, there is increasing evidence showing that designing and adapting educational strategies to cater students’ needs and strengths contributes to higher learning rates, satisfaction, and outcomes. Experiments conducted to assess the effectiveness of personalized instruction have shown that targeted education improves learning efficacy compared to non-targeted teaching practices, leading to greater achievements and performance (Chen et al., 2005; Chen, 2011; Hsu et al., 2013; Kim et al., 2014; Tseng et al., 2008).
Online Personalization and Learning Analytics
Although personalization of learning strategies does not strictly require the use of web technologies, digitalization facilitates systematic collection of large quantities of data as well as generation and delivery of personalized content, particularly in the context of self-learning, professional development, and distant education. Furthermore, e-learning brings two other major advantages: speed and scalability. In addition to providing web-based virtual learning environments (VLEs), digitalization enables fast and automated or semi-automated development of personalized instruction. For example, the analysis of users’ trace data captured through remote devices allows instant creation of real-time individual learner profiles. Those profiles can then be used to customize teaching and improve outcomes. Moreover, technology has enabled the delivery of education at scale through massive open online courses (MOOCs) and small private online courses (SPOCs). Overall, web-technologies have brought innovation to pedagogies and teaching practices to improve learning efficacy.
Advances in information technologies, artificial intelligence, and database systems in the last decades have boosted the use and penetration of personalized education (Garito, 1991; Luckin & Cukurova, 2019). Multiple sources of data have been reported to enable technology-enhanced adaptive learning, including students’ preferences, learning achievements, and learning logs (Xie et al., 2019). The collected information is used to build personalized profiles that comprise individual student performance, learning style, and ability (Chen et al., 2005; Klašnja-Milićević et al., 2011; Schiaffino et al., 2008; Tseng et al., 2008). These profiles constitute the foundation to adapt material difficulty and instructional strategies to specific learner needs and skills. By profiling learners in terms of their engagement and achievements, instructional design can then be implemented to cater for more personalized learning paths.
The development and implementation of online personalized learning systems has been largely facilitated by the emergence of research disciplines such as Educational Data Mining (EDM) (Baker & Yacef, 2009) and learning analytics (LA) (Gašević et al., 2015; Long & Siemens, 2011). EDM and LA share a common goal of understanding the process of learning to improve teaching practices (Siemens & Baker, 2012) and tap on both machine learning methodologies and human judgment to achieve their objectives. Although these disciplines share common objectives, EDM deals with developing methodologies to analyze educational data and discover new insights about how students learn, whereas LA focuses on optimizing learning processes and outcomes. The role of human judgment in these fields is also different. For EDM human judgment is primarily a tool whereas for LA it is an aim (Siemens & Baker, 2012).