On the Cognitive Load of Online Learners With Multi-Level Data Mining

Introduction

As early as 2012, massive open online courses have proliferated. After a decade of development, online learning has begun to take shape. According to relevant statistics, as of November 18, 2013, Coursera has 5.4 million registered users, Udacity has more than 1 million registered users, and EDX has more than 900 thousand registered users (Laura Pappano, 2012). Online learning has become a popular way of learning. Many people are attracted, including researchers, educators, and learners. Until 2020, due to the sudden outbreak of COVID-19, online learning was rapidly propelled to the forefront of the education era. It became an effective method for teachers and learners to take continuous teaching and learning. However, with the continuous expansion of the scale of online learning, there are common phenomena, such as high dropout rates (Kizilcec et al., 2017), low participation rates (Orji et al., 2020), low completion rates (Khalil & Ebner, 2017), and high back accessing rates (Wu et al., 2018) among learners of online learning.

The most interesting phenomenon is learners’ frequently back accessing behavior. Bing Wu and Xiao (2018) found that 15.68% of the accessing activities are back accessing, and each learner has an average of 1.33 times back accessing behavior. So why do learners spend a lot of time on back accessing activities? The direct reason may be that learners have missed some learning resources, or have forgotten important learning content in the process of learning, which makes it hard for learners to continue learning. However, the essential reason may be that learners need to deal with more and more information in online learning. In the process of information processing, the capacity of information work is limited (subject to working memory) (Paas et al., 2010). If the amount of information received by learners exceeds the capacity of working memory, an additional cognitive load will be generated (Sweller et al., 1998), which will reduce the learning efficiency of learners and lead to continuous back accessing behavior. Therefore, it can be speculated cognitive overload of learners may be caused by a large number of learning tasks.

At present, the relationship between learners’ cognitive load and grade has been widely studied (Atiomo, 2020; Kirschner et al., 2011; Tzafilkou et al., 2021). Fairclough et al. (2005) pointed out that when the cognitive load exceeds the total cognitive load of the individual, job performance will decline to some extent. De Waard and Brookhuis (1996) further pointed out that the relationship between task demand and task performance is an “inverse U-curve” relationship.

With the rapid development of online learning, the measurement of online learners’ cognitive load has become an urgent problem in the development of personalized online learning. Based on this, multi-level data mining technology was used in this study to analyze and mine the relationship between grades and online learners’ behaviors layer by layer, using data obtained from the Canvas Network platform. The purpose was to map the specific learning behavior characteristic to the cognitive load by exploring the relationship between learning behavior characteristics and grades. The experimental results show that there is an “inverse U-curve” relationship between grades and “nevents.” Therefore, “nevents” can be mapped to the cognitive load of online learners. The experimental results also show that the special learning rules hidden behind the data can be effectively mined by the multi-level data mining technology.