Structural Gender Differences in LMS Use Patterns among College Students

Abstract

The aim of this study is to investigate male and female college students’ use patterns of a learning management system (LMS) in an e-learning environment. This study evaluated the structural differences between male and female college students in their LMS use patterns through a multifactor model. The research was conducted with 443 participants at a university in Korea. Four factor structures comprising 14 items measured on a five-point Likert scale were used for the analyses. After confirmatory structures for each gender were modified, the equivalence was examined by testing for factorial invariance and the latent means. The results indicated that, for three factors, male students used the LMS more than females and that neither gender preferred communicating and collaborating with each other. It was also found that students understood learning activities in more diverse ways than through theories. The results, which reflected Korea’s general educational context, indicated that a gender digital divide issue remains to be bridged and left recommendations for comprehensive development including the search for strategies for more participative LMS operations.

1. Introduction

The e-learning market is continuously developing. In particular, interest in the sustainability of education through online methods is growing further in the recent 2019 novel coronavirus pandemic. A major factor contributing to the high growth of the e-learning market, which will be worth more than $300 billion by 2025, is the increasing shift toward diverse educational environments such as massive open online courses with technologies powered by artificial intelligence [1]. In higher education, the successful use of a learning management system (LMS) is important for the implementation, satisfaction, management, and sustainable improvement of the quality of learning [2,3]. The diversification of the features of learning management systems (LMS) greatly contributed to the development of e-learning. In this context, constructivist e-learning support systems including participation, exchanging ideas, reflection, and collaborative learning have been recently emphasized alongside the provision of learning materials [4].

As technology continues to evolve, a gap has emerged between benefits that cannot be evenly distributed based on individual backgrounds. The digital divide occurs because of differences in race, age, region, and so on. Among the issues is the view that men are more advantageous and confident in using technology than women [5,6]. Since this could lead to unequal experiences in terms of learning processes and outcomes, the gender divide should be investigated and addressed. However, research on this issue is lacking in Korea, where information technology-related indicators are world-class [7,8], as there are preconceptions that the people are using information and communication technology environments in great abundance.

In this study, we examined the differences in the LMS use patterns of male and female college students in an e-learning environment in Korea. The purpose of the research was twofold: (a) to understand factorial structures of LMS use patterns in male and female students and (b) to analyze the structural differences between genders.

1.1. Learning Management System

An LMS is web-based software using a database to integrate interactive learning environments and administration as well as to facilitate online instruction [9]. With the rapid development of information technology, academic institutions are investing in LMSs to deliver and manage e-learning courses. Blackboard, Moodle, and Canvas are representative commercial LMSs that have been working to improve existing products and to develop more advanced ones. In the case of Korea, domestic e-learning companies are the main suppliers of LMSs, meeting customers’ specific needs [10].

Compared to the limited learning activities of the past, recent e-learning platforms have evolved to deliver authentic and problem-based learning to support the provision of real-life tasks [11]. This has enabled students to interact with complex and realistic problems in collaborative online contexts [4]. Teacher-centered tutorials have long been a major instructional method in Korea; however, interest in and research on constructivist interactions have recently been increasing [12].

1.2. Gender Digital Divide

The digital divide refers to the gap between individuals, households, businesses, and geographic areas at different socioeconomic levels in terms of their opportunities to access information technology [13]. Regarding gender issues pertaining to the divide, American Association of University Women [5] noted that the advent of technology in schools resulted in a discouraging new gap as females were isolated from computer activities. Furthermore, it was found that women’s interest in information and communication technology was declining [14]. The gender divide is a highlighted area of interest in issues related to the digital environment.

After resolving the initial “access” issue, the use of a computer and Internet skills are of interest [15,16]. In other words, the problem now often pertains more to actual computer use than to the haves and have-nots. In addition, social awareness is also critical to understanding the issue. Cooper [17] argues that the gender stereotypes that men dominate computers can constitute a self-fulfilling prophecy, further undermining the divide. A meta-analytical study on gender digital divide showed that research variables on Information and Communications Technology (ICT) general attitudes and satisfaction, ICT confidence, ICT interest and motivation, and actual use of ICT tools and applications were statistically in favor of male students [18]. However, another meta-analysis of the gender differences in students’ ICT literacy contrasted previous results in that the gender differences in ICT literacy were significant, were positive, and favored female students [19].

In Korea, the gender divide is more complicated. Korea is reportedly one of the top innovative countries worldwide in terms of technology and Research and Development (R&D) [7]. In addition, a high standard of living and the world’s best Internet infrastructures position Korea as a testbed to lead the world in developing new educational tools [8]. As the digital environment appears abundant, the interest in the digital divide could be reduced by assuming that the benefits would be evenly distributed.

1.3. LMSs and Gender

Previous studies investigated LMS use patterns in terms of gender. The objectives of these studies were mainly to understand the differences or divide between males and females. Questions generally included gender differences in frequency or preferences for features in LMS use. Many of these studies had no evidence of statistical differences [20,21,22,23]. Meanwhile, Li, Wang, and Campbell [24] found that male students had a significantly higher mean for the frequency of using a course management system than female students. Unfortunately, most of the research regarding this issue conducted t-tests or ANOVAs for statistical methodologies and was limited to identifying structural differences among the sample.

In this study, we used the digital divide to figure out structural gender differences in the use of an LMS in a university in Korea. The reason for exploring factorial invariances of the gender divide was to analyze potential inequality in the core aspects of the design and use of e-learning. The significance of the study is that it was conducted in Korea where not much attention has been paid to the gender divide. Our analysis directly addresses this understudied area.

The research questions were as follows. First, what are the baseline models of each gender that explain the factorial structures of the LMS use patterns? Second, is the factorial structure equivalent across populations? Third, do the latent means of the constructs differ by gender?

2. Materials and Methods

2.1. Research Context and Participants

In this study, data from undergraduate students enrolled in e-learning courses at a private university located in Seoul, Korea were collected in the 2018 fall semester. The survey was conducted at the end of the school year to ensure that students could respond after fully experiencing learning activities using an LMS. Among 1545 students who were taking large-scale e-learning courses and were asked to answer the questionnaire for the study, 448 students voluntarily participated in the survey. After eliminating five data with missing values, 443 valid samples (male = 223 and female = 220) were used for the analyses. Structural equation modeling was employed to examine the research questions, and EQS software was used for the data analyses [25].

2.2. Research Methods

The questionnaire was initially developed based on the LMS use patterns. It was then reviewed by two experts in educational technology and another two experts in educational measurement to enhance the validity of the items. In total, 14 items were formulated based on the features of the LMS used by the university. The features consisted of four factors. First was learning content (LC). Items included “(I) watch learning video clips,” “read learning materials,” and “take open lectures.” Students were supposed to watch the video clips or to read the articles uploaded by instructors and professors. The LMS also offered open lectures for the students to study diverse academic areas. Second was instructional information (II) including items such as “check the syllabus,” “check notices,” and “check my attendance status.” This factor was related to the main points students had to check when they took the class. Third was task performance (TP). The items were “submit assignments,” “take votes,” “answer surveys,” and “take exams.” Students were required to perform these activities to participate in the classes. Last were academic interactions (AI), which included the items “use Q&As,” “participate in group activities,” “hold debates,” and “use bulletin boards.” These aspects aimed to enrich the class. The survey items with related factors are displayed in

Table 1. Survey items for the hypothesized factor structure.

Factor	Variable	Item
F1: Learning Content (LC)	V1	Watch learning video clips
	V2	Read learning materials
	V3	Take open lectures
F2: Instructional Information (II)	V4	Check the syllabus
	V5	Check notices
	V6	Check my attendance status
F3: Task Performance (TP)	V7	Submit assignments
	V8	Take votes
	V9	Answer surveys
	V10	Take exams
F4: Academic Interactions (AI)	V11	Use Q&As
	V12	Participate in group activities
	V13	Hold debates
	V14	Use bulletin boards

.

3. Results

The results of the descriptive analyses of the two models are provided in

Table 2. Results of the descriptive analyses.

Factor/Variable	Male		Female
	M	S.D.	M	S.D.
LC	3.82	1.14	3.77	1.04
V1	4.13	0.96	4.00	0.90
V2	4.16	1.13	4.10	1.04
V3	3.18	1.32	3.21	1.19
II	4.26	0.83	4.34	0.71
V4	4.08	0.95	4.08	0.86
V5	4.33	0.76	4.45	0.62
V6	4.38	0.77	4.48	0.64
TP	4.20	0.96	4.15	0.88
V7	4.46	0.80	4.54	0.62
V8	3.91	1.11	3.84	1.05
V9	3.99	1.03	3.88	1.01
V10	4.45	0.88	4.35	0.83
AI	3.20	1.35	3.09	1.19
V11	3.53	1.30	3.36	1.18
V12	2.91	1.41	2.73	1.27
V13	3.31	1.34	3.27	1.20
V14	3.04	1.36	2.99	1.11

.

A first-order confirmatory factor analysis (CFA) was performed to test the multidimensionality of the theoretical constructs. There were 14 observed variables, and the 4 factors were intercorrelated. Since the kurtosis of the male and female model was more than 5.00 (14.17 and 7.68, respectively), they were regarded as nonnormally distributed. Therefore, the Santorra and Bentler (S-B) χ² and robust statistics portion of the output were used in this study [26].

3.1. Testing for Factorial Validity

The initial tests for the validity of the hypothesized models with the four-factor structure of the LMS use patterns of the male and female student groups were less than optimal for both groups. Using the Langrange Multiplier Test (LM Test), ordered univariate test statistics and the χ² univariate increments associated with the cumulative multivariate statistics were reviewed to test hypotheses bearing on the statistical viability of specified restrictions [27].

Regarding the male student group, the error covariance between items 12 and 13, items 8 and 9, and items 3 and 10 exhibited profound effects. In addition, the cross-loadings including the loading of item 4 and factor 1 and of item 3 and factor 2 contributed to the misfit of the model based on the LM χ² statistic and related probability value. On the other hand, the error covariance between items 7 and 10, items 12 and 13, and items 4 and 5 were found to be the parameters that were contributing substantially to model misfit for female students. The cross-loadings of items 7 and 10 and factor 2, of item 4 and factor 1, and of item 10 and factor 2 were specified as free parameters and were reestimated. Figure 1 displays the baseline models of factorial structures for both genders.

Changing the parameters of both groups substantially improved the model fit statistics from the initially hypothesized model. The newly specified parameters in the baseline models were regarded as significant. The models’ goodness-of-fit statistics were as follows:

• Male students: S-B χ² (66) = 131.713; comparative fit index (CFI) = 0.958, incremental fit index (IFI) = 0.958, root mean square error of approximation (RMSEA) = 0.067, 90% confidence interval (C.I.) 0.050, 0.083
• Female students: S-B χ² (65) = 109.998; CFI = 0.961, IFI = 0.962, RMSEA = 0.056, 90% C.I. 0.037, 0.074

3.2. Testing for Factorial Invariance

With the baseline models for each group, the equivalence of the LMS use patterns was tested. First, configural invariance was tested. As a result, the model’s goodness-of-fit statistics revealed a well-fitting multigroup model with an S-B χ² (131) value of 243.104 (CFI = 0.959, IFI = 0.959, RMSEA = 0.062, and 90% C.I. 0.050, 0.074). Therefore, it could be concluded that the structure of LMS use is optimally represented as a four-factor model, with the pattern of factor loadings specified in accordance with the initial multigroup model.

Next, the question on equality with respect to the measurement model focusing on factor loadings and measurement error variance-covariances was addressed. Thus, equality constraints such as one cross-loading (V4 and F1) and one error covariance (E13 and E12) were commonly specified for each group. Review of the multigroup model underwent deterioration in model fit (corrected ΔS-B χ² = 326.646, p < 0.001; ΔCFI = 0.103) and exhibited not a good fit to the data (CFI = 0.856; RMSEA = 0.108; 90% CI = 0.098, 0.118). This was probably because the other freely estimated cross-loadings and error covariances used in the baseline model for each group were not considered in this measurement model. The LM test statistics revealed one commonly specified error covariance (E13 and E12) to be noninvariant (p < 0.01) across the two student groups. This suggested that the male and female students interpreted the item content differently.

In addition, the results of testing for the invariance of the factor covariances indicated significant differences in ΔS-B χ² values (corrected ΔS-B χ² (25) = 70.65) and that the difference in CFI was not negligible (ΔCFI = 0.107). The LM Test also showed one error covariance (E13 and E12) with a probability of <0.05. The error covariance associated with the items is one component of the model; however, the invariant requirement of errors across groups is considered excessively stringent and of little practical value [28]. Therefore, no further interpretation of the nonequivalence with the error covariance was conducted. The results of the invariance tests are summarized in

Table 3. Results of the invariance tests for males and females.

Model	S-B χ2	df	CFI	RMSEA	ΔS-B χ2a	p
Configural	243.014	131	0.959	0.062
Measurement	538.373	150	0.856	0.108	326.646	<0.001
Structural	556.959	156	0.852	0.108	70.65	<0.001

.

3.3. Testing for Latent Mean Differences

The equivalency of the latent mean structures of the four factor dimensions LC, II, TP, and AI was tested for male and female participants. Given that the female group was designated as the reference group, the estimates were related to the males. The analyses were based on robust statistics; thus, the estimates were the robust standard errors and the resulting z-statistics. The results are provided in

Table 4. Results for latent mean differences.

Construct Equations	S. E.	Z
F1 = 1.427 * V999 + 1.000D1	0.043	33.405 *
F2 = 0.467 * V999 + 1.000D2	0.629	0.743
F3 = 2.102 * V999 + 1.000D3	0.066	31.637 *
F4 = 1.599 * V999 + 1.000D4	0.033	48.291 *

V999: Constant, D: Disturbance * p < 0.05.

.

4. Discussion and Conclusions

This study evaluated the structural differences between male and female college students in their LMS use patterns through a multifactor model with a latent mean differences analysis. Through the research, the authors aimed to understand how issues pertaining to the gender digital divide would appear to each gender’s preferences for using the features of an LMS. To this end, 443 undergraduate students enrolled in e-learning courses at a university in Korea were asked to complete the questionnaire on LMS usage types using a five-point Likert scale. This study was conducted in a specific country for a relatively small number of college students, so there is a limit to generalization. Survey items for the factor structure could also be changed in other environments. However, the research is meaningful in that it systematically analyzed gender-specific usage patterns for sustainable use of the LMS.

The results of the latent mean differences testing showed that the males used three of the four factors more actively than the females. Specifically, male students used the system more than females to study, to perform tasks, and to talk/listen to others. Rather, no statistical difference was found for the instructional information factor, suggesting that males and females used the LMS in a similar way to get class-related information such as the syllabus, instructors’ notices, and attendance status. In addition, both males and females did the most learning activities for this factor (4.26 and 4.34 on a 5-point Likert scale, respectively).

The academic interactions factor scored lowest, indicating that the students did not use the LMS much for informal learning. Specifically, the responses of both males and females were less than three points (2.91 and 2.73 each) for the item “participate in group activities.” This means that instructors probably did not provide group activities and that the students did not perform collaborative work. This result was somewhat expected in the Korean learning environment. In Korean education, teacher-oriented directive classes have been held for a long time and students seldom ask questions or discuss subjects voluntarily [29,30]. Recently, a movement for change has emerged, emphasizing the importance of learner-centered instruction and collaborative learning [31,32].

The factorial invariance test yielded impressive results. The baseline models of the factor structure were formulated in a relatively complicated way. For male students, group activities and debates, votes and surveys, and open lectures and exam taking were highly suggestive of content overlap. In addition, they seemed to think that checking the syllabus was part of learning. For females, group activities and debates, syllabi and notices, and assignments and exams were considered related. Moreover, they thought assignments and exams were part of instructional information and that syllabus checking was part of learning. The results suggested that students understood learning activities in more diverse ways compared to the categories provided in the study. Even though most of the additional covariances and factor loadings for better model fit could be interpretable as stated in consideration of LMS contexts and environments, the reason male students related open lectures with exams should be investigated in more depth. The results for the non-invariance of the commonly specified error covariance were not interpreted because of their stringency and impracticality.

Based on the research, the following recommendations are made for the sustainable improvement of online learning environments. Especially, the importance of sustainability in education is being emphasized at a time when demand for online learning is expected to gradually increase in line with rapidly changing social phenomena such as the development of science and technology and the quarantines caused by infectious diseases. First, this study showed that males more often used three of the four features of the LMS than females. If this is for no other reason but for reasons that men are more familiar with electronic devices and use technology more than women, it is in line with the stereotype. Since this study had not been conducted in a developing country with poor access to technology and where females experience inequities, the implications of the results are more complicated. Given that this research was conducted in Korea, a leader in the IT industry, the result implied that the gap could still exist. Moussa and Seraphim [33] argued that there was little evidence to suggest that new technologies alone enable women to alter male-dominated gender power relations in society. Rai [34] also pointed out that women in some sections in South Asian countries did not have equal status and rarely participated in the decision-making processes at the household level, despite the improved development and use of science and technology. What does it mean to have a difference in actual use in an e-learning environment where there is no difference in technology access? It is required to research further to figure out the factors for females not to use technology well in rich IT environments.

Second, the results revealed that students still did not use the LMS much to ask questions, to participate in group activities, to hold debates, and to use bulletin boards. Compared to the first suggestion that was related to the gender imbalance in use, this is about the result that students hardly interact online, regardless of gender. This should go beyond the simple digital divide to look at educational and cultural backgrounds. Students’ passiveness in a class is aligned with the problems of Korean education, as pointed out. Students may feel less pressure to speak online, making it necessary to emphasize collaborative learning through the LMS. Therefore, optimal instructional strategies should be applied and features with high usability should be provided to ensure that instructors and students participate in online activities. The instructors’ LMS usage patterns should also be considered [35]. Such an attempt will be an important starting point for transforming education in Korea.

Last, the comprehensive development of LMS is expected. In particular, various dynamic features such as virtual reality, augmented reality, and artificial intelligence will be introduced to improve learning effectiveness as technologies rapidly evolve. This will catapult both online and offline education to higher levels of excellence as well as will bridge digital divide issues including gender gaps. In this context, it is important to design instructional environments that enable students to participate and effectively use an LMS.