Determinants of College Students’ Online Fragmented Learning Effect: An Analysis of Teaching Courses on Scientific Research Software on the Bilibili Platform

Abstract

In the era of mobile Internet, online fragmented learning has become one of the mainstream ways for college students to study independently. Analyzing the online fragmented learning effect (OFLE) has essential reference value for guiding the online learning behavior of college students and optimizing the allocation of online learning platform resources. It also provides sustainable solutions for enhancing college students’ online independent learning effect. For this paper, we took college students who have studied teaching courses on scientific research software (TCSRS) on Bilibili as our research subjects, constructed a theoretical model of college students’ OFLE based on the theory of online learning and fragmented learning, verified the model through using questionnaire data, and analyzed the determinants of college students’ OFLE and its mechanisms. Our results indicate the following: (1) Most college students spend a dispersed amount of time learning TCSRS, showing the characteristics of “fragmentation”. (2) Learning motivation (LM), self-efficacy (SE), and fragmented time utilization (FTU) have a significant positive effect on college students’ OFLE. (3) Knowledge fragmentation (KF) affects college students’ OFLE, but it is not significant. These dimensions provide a theoretical reference for quantitative research on the impact of college students’ OFLE. Finally, also in this paper, improvement countermeasures are proposed from the perspective of improving college students’ OFLE and ability.

1. Introduction

With the popularization of mobile intelligent terminals, the updating and application of Internet communication, computer networks, and other technologies, a variety of online learning platforms, course resources, and learning modes have continued to emerge, prompting significant changes in students’ learning habits and behaviors and effectively enhancing their independent learning ability [1]. Online learning gradually eliminates the need for a wired network and fixed terminal equipment, allowing learners to enter the learning state at anytime and anywhere, making their learning arrangement more and more random and uncertain; thus, online fragmented learning comes into being [2]. It is worth noting that during the COVID-19 pandemic, offline education in countries all over the world was greatly affected, and many educational courses had to be switched from offline to online. Since the COVID-19 pandemic, online learning has been accepted and trusted by the majority of students. They hope to make full use of their spare time to consistently learn knowledge on online learning platforms in today’s era, when knowledge is more scattered and disordered, and online fragmented learning has subsequently become an essential supplement to the traditional learning methods of Chinese college students [3]. Therefore, in order to ensure the sustainability of online learning, it is necessary to study students’ OFLE when studying on online learning platforms.

Chinese college students are required to write course papers and graduation papers during their studies, which means they need to have a certain degree of scientific research literacy and support when completing their studies. However, on the one hand, there is a general mismatch between the content of TCSRS and the skills that college students need to master at present, making it difficult for most college students to master the standard and fragmented methods of research software in the courses. On the other hand, most Chinese students have more daily courses and participate in many social practices, such as clubbing on weekends, so their learning time is scattered, leading to a lack of concentration and poor learning effects. This requires college students to dedicate more of their spare time to learning using scientific research software. Given this, Bilibili and other online platforms have aimed to fully consider the learning characteristics of college students, providing them with rich online courses and enticing many college students to learn via their platforms. Now, Bilibili has become an essential channel for Chinese college students looking for TCSRS, as well as a significant platform for promoting online fragmented learning among Chinese college students.

Existing research shows that many factors influence the sustainability of online learning and its effects. Students’ learning attitudes, acquired knowledge, and skills play an essential role in solving sustainable education problems. However, they are faced with the problem of how to translate these competencies into sustainable educational practices [4]. Motivation is a prerequisite to realizing sustainable learning activities, and sustainable motivation is a necessary factor for engaging in the learning process and achieving learning goals [5]. SE plays a vital role in promoting sustainability in students’ online learning environments and their learning outcomes [6]. In addition, the more one uses their spare time to learn a broader range of content, the easier it is to expand students’ cognitive profiles [7], which is in line with the learning characteristics of research software. At present, Chinese college students’ OFLE could be better, and it is easily disrupted by various determinants [8]. So, what are the determinants that influence the OFLE? How about its mechanisms? It is a real problem that needs to be solved urgently? We aimed to answer these questions and explore what the determinants of college students’ OFLE are and how these determinants affect OFLE. This study is of great practical significance for achieving the sustainability of online fragmented learning among college students in guiding them to change their online fragmented learning behaviors, reasonably use their fragmented time, and enhance their OFLE. It also provides a basis for online learning platforms to optimize resource allocation according to college students’ learning characteristics.

Different from traditional learning, online fragmented learning using cell phones and other digital tools as learning carriers is more flexible, with the characteristics of networked learning tools and fragmented learning content, which can make up for the shortcomings of traditional classroom-based learning and improve students’ learning efficiency [9,10,11]. Correspondingly, Aldholay et al. [12] argue that online fragmented learning allows learners to access learning websites anytime and anywhere, which can help learners make better use of their fragmented time. However, Tang et al. [13] believe that although online fragmented learning has specific feasibility, it is still in its infancy and in need of further optimization. Judging from the existing literature, the existing academic research on online fragmented learning mainly focuses on constructing a fragmented learning platform, analyzing fragmented learning advantages and disadvantages, and designing fragmented instructions. However, few studies have explored the learning effect, especially the effects of LM, SE, KF, and FTU on OFLE and their mechanisms. Based on the above four dimensions, this paper proposes a new model to verify the determinants of OFLE; explains the effects of LM, SE, and FTU on online fragmented learning; provides a theoretical basis for understanding online fragmented learning and sustainable learning; and expands the research scope of online fragmented learning.

The remainder of this paper is structured as follows: This paper first defines the variables through presenting a literature review in Section 2 and then constructs a theoretical model combining online and fragmented learning research in Section 3. College students who have studied TSCRS regarding SPSS, Amos, Stata, and Origin on Bilibili were selected as the research subjects for this study (more information on the research subjects is provided in Section 4), and they helped to verify the aforementioned model through providing questionnaire data, which helped us explore the determinants of college students’ OFLE and its mechanisms, as discussed in Section 5 and Section 6. Finally, a series of countermeasures to enhance college students’ OFLE in a sustainable way by optimizing the resource allocation of online learning platforms and changing online learning behaviors are proposed in Section 7.

2. Review of Related Studies in the Literature

2.1. Research Status of Online Fragmented Learning

As early as the 1980s, the concept of “fragmentation” appeared in research on reading behavior, and now it is mainly applied to education and other fields [7,14]. At present, scholars are still yet to construct a clear definition of fragmented learning. Yang et al. [8] point out that fragmented learning is an informal learning method that divides learning time and content so that learners can use fragmented time to learn fragmented content. Online fragmented learning divides learning content so that learners can use the Internet platforms to learn at any time and anywhere.

In recent years, with the rapid development of information technology, “Internet + fragmented learning” has ushered in many development opportunities, and online fragmented learning has become the key to the development of education informatization in China [15]. Online fragmented learning depends entirely on learners’ interests, goals, thinking, and other needs. It breaks the time and space limitations of traditional learning, enables learners to use their time to select the content fragments they are interested in on the online platform for learning, and can effectively help learners diffuse their thinking [15,16]. However, due to the mixed quality of network information and the fact that fragmented learning can easily lead to distraction and inattention, fragmented learning can only be a helpful supplement to systematic learning [16]. Song et al. [17] confirmed the critical role played by online fragmented learning from the perspective of the mediating role of the Internet, which can enhance the learning effect.

2.2. Research Status of Research on Online Learning Effective and Determinants

With the widespread popularization of mobile electronic devices and the rapid development of the education industry, online learning has become one of the essential ways for learners to learn. Online learning has become popular in higher education because of its facilitation of a flexible learning style, the high selectivity of learning content, and the fact that it allows for controllable learning progress. In recent years, scholars have researched online learning’s effects and determinants. For example, Chang et al. [18] constructed an impact model for the performance of college students taking an online English course based on social cognition and social cognitive expectation theory and proved that students’ emotional intelligence is affected by LM and SE, which could indirectly affect their learning performance. Based on the grounded theory, Li et al. [19] studied learning expectations, learning quality, learning interaction, learning performance, and their combined impacts on students’ online learning during COVID-19. Accordingly, Song et al. [17] put forward a structural equation model (SEM) of the fragmented learning effect, confirming that LM, learning resources, learning input, and the Internet can affect the fragmented learning effects, in which the Internet has a mediating effect. In addition, Alismaiel et al. [20] believe that students’ interactions with peers and teachers affects students’ online learning.

2.3. Research Progress on LM and SE

Motivation is the driving force that determines people’s intrinsic behavior and serves as a catalyst for improving individual work performance and efficiency [21]. LM is the driving force that drives students to learn, and it is used to encourage them to achieve goals [22]. Scholars have studied the impact of LM and online learning effects. For example, Lin et al. [23] investigated students’ online language learning strategies and confirmed that more learning strategies can help learners improve their academic performance. However, low LM will hinder learners from making online learning strategies. Correspondingly, Tokan et al. [24] confirmed that intrinsic and extrinsic LM directly affect students’ academic performance.

SE is an aspect of human self-perception that determines the degree of effort an individual puts in to achieve a goal and influences the realization of personal goals [22,25]. In online learning, SE is the user’s self-perception of their behavioral ability to complete learning tasks using a particular platform. SE can help students strengthen their determination to complete a task and affect their learning behaviors, reflecting positive performance, which enhances academic performance [26,27]. Bandura [28] argues that students with low SE tend to avoid complex tasks because they perceive the task as a threat to them. In addition, Kustyarini [29] used structural equations to demonstrate that SE mediates learning outcomes. In other words, to a certain extent, SE plays a decisive role in learners’ self-confidence in online learning.

2.4. Research Status of Research on the Relationship between Fragmented Learning and Learning Effects

In the Internet information age, fragmented knowledge is updated quickly. If students can reasonably utilize fragmented knowledge, it will help them to improve their knowledge system structure [30]. In recent years, scholars have explored fragmented learning and its effect on learning. For example, based on using the BPNN algorithm and studying genetics, Yang [31] found that English fragmented learning has a better effect in the mobile information environment but also found that online fragmented learning has disadvantages, such as knowledge dispersion and a lack of system structure. However, Li et al. [32] believe that although fragmented learning provides more possibilities for online learning, it is still necessary to ensure that all the contents of a certain knowledge point are learned over a long time to guarantee the online learning effect. In addition, LM [8,17,22], SE [33,34], and other determinants will have an impact on the fragmented learning effect. Although some scholars have explored the impact of college students’ fragmented reading on their cognitive development, from content fragmentation, time fragmentation, and attention fragmentation [7], they have not been used in the study of online fragmented learning with respect to scientific research software; also, the determinants of OFLE have not been explored, nor have the roles of the determinants of OFLE. Hence, investigating these aspects was the theoretical basis of this study.

3. Research Model and Hypotheses

3.1. Model Assumptions

Based on the above theoretical hypotheses, combined with online learning and fragmented learning theory, this paper draws on Song et al.’s [17,18] SEM of the online learning effect to construct a theoretical research model (Figure 1).

Figure 1. Proposed model.

3.2. Model Hypotheses

LM assesses students’ confidence in their ability to persist in effective learning [35], which drives students to learn and can affect their academic performance [22,24]. Because of this, this paper proposes Hypothesis 1:

Hypothesis 1. 

LM has a significant positive effect on college students’ OFLE.

SE refers to people’s judgment of their ability to perform a particular task [26,36]. College students with stronger SE tend to have a stronger belief in themselves, which can help them accomplish their learning tasks and improve their academic performance [27]. Liang et al. [37] point out that students with stronger SE prefer the Internet learning environment and have a stronger relationship with online learning. Therefore, this paper proposes the following hypothesis based on SE:

Hypothesis 2. 

SE has a significant positive effect on college students’ OFLE.

In the mobile Internet era, which relies on network-based platforms, complicated knowledge is divided into several independent parts, characterized by “knowledge fragmentation” [38]. College students are affected by the school curriculum, and their time is divided into several irregular segments, characterized by “time fragmentation” [39]. In order to improve the learning ability of college students, they can use fragmented time to find fragmented content for decentralized learning, which breaks the time and space limitations of traditional learning and provides more possibilities for online learning [31]. Therefore, this paper proposes the following hypotheses based on KF and FTU:

Hypothesis 3. 

KF has a significant positive effect on college students’ OFLE.

Hypothesis 4. 

FTU has a significant positive effect on college students’ OFLE.

4. Materials and Methods

4.1. Data Collection

SPSS, Amos, Stata, and Origin are the essential scientific research tools for researchers when conducting empirical analyses and drawing graphs, and both of these tasks require college students to have the essential ability to independently use the above four software to complete data analysis and presentation. Therefore, after the reliability of the pre-survey was validated, we adopted a random sampling technique and selected college students who have studied courses on using SPSS, Amos, Stata, and Origin on Bilibili as the research subjects and distributed questionnaires to them. The use of the random sampling method led to the selection of 596 college students who have taken the above TCSRS on Bilibili, and these students were subsequently surveyed. Thirty-three questionnaire responses were excluded from the questionnaire data analysis due to them answering the questions too quickly, their answers being too extreme, and/or their answers for all options being the same, leaving 563 valid questionnaire responses, with an effective questionnaire recovery rate of 94.46%.

4.2. Participants

This study’s participants were college students who had used Bilibili for TCSRS. Table 1 shows the demographics of all participants. Of the 563 research subjects, there were 269 males and 294 females, meaning the male/female ratio was close to 1:1. Regarding the educational level of the interviewees, the participants were mostly undergraduates (60.39%) and master students (27.53%). It is assumed that undergraduate and master students, influenced by daily study, are more willing to use fragmented learning time to learn TCSRS on fragmented learning platforms (e.g., Bilibili) to improve their research literacy and ability.

Table 1. Demographic statistics of participants.

Gender	Educational Level
	Junior College Student	Undergraduate	Master Student	Doctor	Total
Male	28 (10.41%)	156 (54.99%)	75 (27.88%)	10 (3.72%)	269 (47.78%)
Female	22 (7.48%)	184 (62.59%)	80 (27.21%)	8 (2.72%)	294 (52.22%)
Total	50 (8.88%)	340 (60.39%)	155 (27.53%)	18 (3.20%)	563

4.3. Research Survey

Based on the four hypotheses described in Section 3.2, drawing on relevant results from the existing literature on the design of learning effect scales [22,30,40] and combining these with the characteristics of TCSRS on Bilibili and the actual situation of college students’ online learning, we designed the Questionnaire on College Students’ Fragmented Learning of Scientific Research Software on Bilibili. In addition to the demographic information of the participants and the frequency of fragmented learning TCSRS on Bilibili, this questionnaire also includes measurement scales for five research variables, namely LM, SE, KF, FTU, and OFLE, for a total of 19 items. The scale design dimensions and the operationalized definitions of the variables are shown in Table 2. The options for each item of this scale were designed according to a five-point Likert scale, with scores 1 to 5 corresponding to a range spanning from strongly disagree to strongly agree, and respondents could only select 1 score for each item.

Table 2. Operationalized definitions of scale dimensions and variables.

Variables	Items	References
LM	LM1: I hope to use my fragmented time to acquire fragmented knowledge of scientific software on Bilibili.	Rafiola et al. [22] Tang et al. [13]
	LM2: I hope to improve my academic skills through fragmented learning software teaching courses on Bilibili.
	LM3: I hope to learn to master the relevant research software knowledge points and use them in practice on Bilibili.
	LM: I hope to do my studies well without being supervised.
SE	SE1: I am able to keep up with the online research software course on Bilibili.	Rafiola et al. [22] Sun [41]
	SE2: I am able to find the solution to the problem on Bilibili.
	SE3: I am confident that I can solve real problems in research through fragmented learning on Bilibili.
	SE4: I can make reasonable use of fragmented time to study effectively on Bilibili.
KF	KF1: I think the content of the research software course on Bilibili is compact.	Zhou [30] Sun [41]
	KF2: I think the coherence of scientific research software courses on Bilibili is very weak.
	KF3: I think fragmented knowledge has some depth.
	KF4: I think fragmented knowledge is comprehensive and rigorous.
FTU	FTU1: I can use the fragmented time to study scientific research software teaching courses on Bilibili.	Zhou [30] Yang et al. [42]
	FTU2: I can flexibly apply fragmented time to the learning of individual fragmented knowledge.
	TFU3: I am able to use fragmented time to sort through or master fragmented knowledge.
OFLE	OFLE: The scientific research courses on Bilibili have further deepened my mastery of scientific research knowledge.	Guo et al. [40] Yang et al. [42]
	OFLE 2: My scientific research practice ability has been improved by fragmented learning of scientific research knowledge on Bilibili.
	OFLE 3: Learning scientific research knowledge online on Bilibili can help me improve my ability to discover and solve problems in my studies.
	OFLE 4: I can learn more scientific research software through the fragmentation on Bilibili.

4.4. Data Analysis

Firstly, we conducted a descriptive statistical analysis of the variables and exploratory factor analysis (EFA) of each dimension using SPSS. Subsequently, a confirmatory factor analysis (CFA) was conducted using Amos to test the combinatorial reliability (CR) and average variance extracted (AVE) of the theoretical model and to verify its goodness of fit. Next, the correlations of the variables were analyzed using SPSS. Finally, a complete SEM of college students’ OFLE was constructed using Amos, and the influence of each variable on OFLE, along with each variable’s mechanism, was tested.

5. Results

5.1. Descriptive Analysis

5.1.1. Descriptive Analysis of Participants

The frequency of college students using Bilibili to learn fragmented information in TCSRS is illustrated in Figure 2. A total of 13.32% of college students learned TCSRS on Bilibili every day, 51.51% of college students did so 2–3 times a week, 24.33% of college students did so only 2–3 times a month, and the remaining 10.83% of college students did so only once a month or more. More than 80% of college students did not study TCSRS on Bilibili at a fixed time, and their learning time was fragmented.

Figure 2. College students fragmented learning TCSRS frequency ratio on Bilibili.

5.1.2. Descriptive Analysis of Variables

The result of our normality test for the descriptive statistics of the variables are shown in Table 3. The mean values of LM, SE, TFU, and OFLE are above 3.5, and the mean value of KF is below 3. The absolute value of the skewness of each variable is within 3, and the total value of the kurtosis is within 8. The fundamental values of the skewness and kurtosis are within the standard range, which means that the data of this scale meet the requirement of approximate normal distribution.

Table 3. Results of our normality test for the descriptive statistics and measurement question items for each dimension.

Items	Average	Standard	Skewness	Kurtosi	Overall Average	Overall Standard
LM1	3.90	0.969	−1.129	1.430	3.920	0.844
LM2	3.93	1.047	−1.108	0.886
LM3	3.92	1.017	−1.014	0.701
LM4	3.93	1.021	−0.876	0.208
SE1	3.70	1.013	−0.920	0.687	3.754	0.804
SE2	3.79	1.038	−0.938	0.679
SE3	3.71	0.985	−0.793	0.511
SE4	3.81	0.994	−0.983	0.864
KF1	2.58	1.186	0.456	−0.701	2.577	1.100
KF2	2.65	1.329	0.472	−0.983
KF3	2.63	1.302	0.351	−1.074
KF4	2.44	1.248	0.630	−0.633
TFU1	3.70	1.039	−0.756	0.169	3.683	0.868
TFU2	3.65	1.070	−0.806	0.163
TFU3	3.70	0.986	−0.807	0.494
OFLE1	3.93	0.934	−0.880	0.672	3.899	0.781
OFLE2	3.84	0.962	−0.946	0.926
OFLE3	3.88	1.035	−1.048	0.784
OFLE4	3.93	0.925	−1.197	1.754

5.2. EFA Analysis

The Bartlett sphericity test value of the sample data in this paper is 5042.269, the KMO value is 0.882, and the significance level is 0.000, indicating that the questionnaire data can explain the relationship between the variables and satisfy the conditions of factor analysis. This paper presents the results of an exploratory factor analysis of the structure of college students’ OFLE, which was carried out using SPSS with Kaiser’s normalized maximum variance approach, and the factor loadings of all items are greater than 0.5, with a cumulative variance explained of 69.338%. The factor loading coefficients of each item and the components are shown in Table 4. It can be seen that the scale has high convergent validity, and this scale can measure college students’ OFLE.

Table 4. Rotated factor loading matrix for exploratory factor analysis.

Items	Factor 1	Factor 2	Factor 3	Factor 4	Factor 5
	LM	SE	KF	FTU	OFLE
LM1	0.808
LM2	0.774
LM3	0.731
LM4	0.711
SE1		0.774
SE2		0.716
SE3		0.712
SE4		0.744
KF1			0.852
KF2			0.872
KF3			0.858
KF4			0.843
FTU1				0.848
FTU2				0.807
FTU3				0.837
OFLE 1					0.747
OFLE 2					0.752
OFLE 3					0.814
OFLE4					0.641

5.3. CFA Analysis

5.3.1. Reliability and Validity Tests

The reliability analysis of the finalized scale data using SPSS found that the Cronbachs’ alpha of the questionnaire is 0.777, which is good enough for subsequent analyses. Table 5 shows each dimension’s CR values and AVE values. The CR value of each dimension is greater than 0.7, and the AVE value is greater than 0.5, which indicates that the model has good reliability and convergence [43,44].

Table 5. Confirmatory factor analysis and test results.

Variables	CR	AVE	Problems	Factor Loading Factor
LM	0.853	0.593	LM1	0.744
			LM2	0.799
			LM3	0.815
			LM4	0.718
SE	0.811	0.519	SE1	0.781
			SE2	0.750
			SE3	0.662
			SE4	0.681
KF	0.811	0.519	KF1	0.789
			KF2	0.845
			KF3	0.831
			KF4	0.813
FTU	0.794	0.562	FTU1	0.771
			FTU2	0.708
			FTU3	0.769
OFLE	0.828	0.547	OFLE1	0.707
			OFLE2	0.674
			OFLE3	0.837
			OFLE4	0.731

In addition, according to Fornell et al. [45], the square root of the AVE is greater than the correlation coefficient to judge the validity. Table 6 shows that the standardized correlation coefficient between any two variables in this study is less than the square root of the corresponding AVE values. Therefore, the discriminant validity of the dimension is suitable.

Table 6. Discriminant validity test results.

Variables	LM	SE	KF	FTU	OFLE
LM	0.770
SE	0.694	0.720
KF	−0.315	−0.259	0.820
FTU	0.220	0.249	−0.049	0.750
OFLE	0.669	0.630	−0.271	0.290	0.740

Note: The diagonal bolded numbers are the AVE square root.

5.3.2. Goodness of Fit of the Model

Table 7 shows the model’s goodness of fit metrics. Referring to the model fitting goodness of fit criteria proposed by Malabika [43], the χ²/df is 2.045, the RMSEA is 0.043, and the CFI, GFI, and AGFI are all greater than 0.9, belonging to the excellent range. The results show that the college students’ OFLE of the TCSRS model on Bilibili has reached a desirable level.

Table 7. Goodness of fit metrics of the model.

Indicators	χ2/df	RMSEA	CFI	GFI	AGFI
Results	2.045	0.043	0.970	0.949	0.932

5.4. Correlation Analysis

The variance inflation factor (VIF) of each variable was less than 5, and there was no multicollinearity, meaning that correlation analysis could be performed [46]. The results of the personal correlation analysis of each dimension are shown in Table 8. It can be seen that (1) LM shows a significant positive correlation with SE, FTU, and OFLE at 0.01, with correlation coefficients of 0.578, 0.186, and 0.571, and a negative correlation with KF at the 0.01 level, with a correlation coefficient of −0.275. (2) SE shows a significant positive correlation with FTU and OFLE at the 0.01 level, with correlation coefficients of 0.203 and 0.531, respectively. However, SE has a negative correlation with KF at the 0.01 level, with a correlation coefficient of −0.220. (3) KF has a significant negative relationship with OFLE at 0.01, with a correlation coefficient of −0.238, and a non-significant negative correlation with TFU. (4) FTU has a significant positive correlation with OFLE at 0.01, with a coefficient of 0.242.

Table 8. Pearson correlation results.

Variables	LM	SE	KF	FTU	OFLE
LM	1
SE	0.578 **	1
KF	−0.275 **	−0.220 **	1
FTU	0.186 **	0.203 **	−0.044	1
OFLE	0.571 **	0.531 **	−0.238 **	0.242 **	1

Note: ** indicates a significance level of 0.01 (two-tailed).

5.5. Hypotheses Testing

5.5.1. Structural Model Building

In order to explore the mechanisms of the determinants of college students’ OFLE, we used Amos for structural equation modeling to draw the structure of college students’ OFLE standardized coefficient model, as shown in Figure 3.

Figure 3. Structure of the standardized coefficient model for college students’ OFLE.

5.5.2. Model Test Results

According to the results of the structural equation model path test for college students’ OFLE on Bilibili (Table 9), it can be seen that (1) LM has significance with OFLE (β = 0.631, p < 0.001), so H1 is valid. (2) SE can positively predict OFLE significantly (β = 0.427, p < 0.001), so H2 is valid. (3) KF is not significant in predicting OFLE (β = −0.084, p > 0.01), so H3 is invalid. (4) FTU can significantly and positively predict OFLE (β = 0.200, p < 0.001), so H4 is valid.

Table 9. Hypotheses testing results of college students’ OFLE.

Hypothesis	Path	Standardized Regression Weights	S.E.	t	p	Results
Hypothesis 1	OFLE <--- LM	0.631	0.051	8.636	***	Valid
Hypothesis 2	OFLE <--- SE	0.427	0.050	6.427	***	Valid
Hypothesis 3	OFLE <--- KF	−0.084	0.017	−2.429	0.015	Invalid
Hypothesis 4	OFLE <--- FTU	0.200	0.026	5.120	***	Valid

Note: *** indicates a significance level of 0.01.

6. Discussion

6.1. Discussion of Model Variables and Values

Through questionnaire data analysis, we verified the college students’ OFLE model. There are innovations in this model. Firstly, compared with the models constructed by other scholars based on the influence of LM, SE, and learning strategy on students’ performance [34], we selected LM, SE, KF, and FTU to reflect the determinants affecting college students’ OFLE, particularly highlighting the “fragmentation” characteristic of online fragmented learning and expanding the model’s explanatory ability and scope for OFLE. Secondly, this model studies the determinants of college students’ OFLE and their mechanisms, extending the theoretical results of online fragmented learning research. It provides academic references for guidance countermeasures to improve college students’ online fragmented learning behavior and enhance the effect of online fragmented self-study, which can help to enhance the interest in and quality of college students’ sustainable online learning. This is one of the key academic contributions of this paper.

6.2. The Relationship between LM, SE, and OFLE

The above results show that LM and SE significantly positively affect college students’ OFLE, and this is consistent with the findings of Bai et al. [47], which prove that LM and SE have a more significant impact on the learning effect in blended and fragmented learning. LM is the “catalyst” for college students’ autonomous learning. It can improve college students’ learning concentration and stimulate their enthusiasm for continuous learning, thereby continuously strengthening the sustainable online learning effect [22]. SE reflects college students’ self-cognition and self-affirmation and is the driving force of college students’ independent learning. If students can always face learning tasks with higher SE, the learning effect of college students will be continuously improved [26]. Putra et al. [48] believe that LM can determine the direction or goal of learning actions and that it is an essential factor in helping students achieve good learning outcomes. In addition, they believe that SE, as an essential factor in self-regulation, influences the level of learning to reach desired learning achievements. Accordingly, Kustyarini [29] suggests that students with more robust SE may have a better learning effect. In summary, whether students can maintain high LM and SE for a long time is crucial to the sustainability of their online learning.

6.3. The Relationship between KF and OFLE

The results of this paper denote that KF does not significantly impact OFLE. Liu et al. [7] showed that fragmented reading significantly negatively affects cognitive depth, which may prevent students from comprehending the texts and information they have read. However, we did not find that KF has a significant effect on OFLE. Knowledge itself needs to be more organized and cohesive. Although research software teaching videos are different from long texts, being more vivid and graphic, and have short durations and the characteristics of fragmented knowledge, it is beneficial for students to watch videos more intuitively and absorb knowledge. “Fragmentation” has changed the original knowledge structure, making it difficult to gain complete and comprehensive knowledge. As long as students organize and process their knowledge in time, they will be able to fulfil their learning goals [49]. Fragmented knowledge is not conducive to college students’ systematic learning, and it is difficult for students to adequately accumulate knowledge [16], resulting in unsatisfactory OFLE, which may be the reason why the impact of KF on OFLE is not significant.

6.4. The Relationship between FTU and OFLE

Research shows that FTU positively affects college students’ OFLE. Zhou [30] believes that college students’ willingness to use fragmented time for English learning can stimulate their subjective initiation in learning, thus improving their fragmented English learning effect. This paper draws a consistent view of college students’ fragmented learning TCSRS on Bilibili. On the one hand, college students are affected by daily course tasks and social activities; thus, their time is inherently fragmented. On the other hand, TCSRS is more dispersed, and concise course videos are easy to understand and access, prompting them to become easily accepted by college students. At the same time, the fragmented learning time breaks the limitations of original online learning and makes learning truly autonomous [49,50]. Online fragmented learning provides opportunities for college students to use their fragmented time. In this case, if college students can make reasonable use of their fragmented time to learn online and expand the scope of their learning over the long term, their OFLE will continuously improve.

6.5. Limitations and Future Research

Although this paper thoroughly studies the determinants of college students’ OFLE and its mechanisms, it still has some shortcomings. Firstly, this study’s sample is insufficient. Only 563 sample responses were included in our study; the results of this study have a certain degree of chance in them, and the quality of the data also needs improvement. Our upcoming study is expected to include more sample data to verify the OFLE model and its determinants. Secondly, only four variables—LM, SE, KF, and FTU—were considered in our study. However, in fact, determinants such as platform characteristics, Internet technology, and learning attention may affect students’ OFLE. Therefore, we will consider including more determinants and considering platform characteristics as the moderating variables in our upcoming study to investigate sustainable online learning in order to better reflect the attributes of fragmented learning on different platforms and its impact on the sustainability of online learning.

7. Conclusions and Improvement Countermeasures

Fragmented learning is an essential trend in the context of the rapid development of the Internet. It has gradually become an indispensable and vital way for college students to acquire knowledge [51]. For this paper, we investigated 563 college students who have studied SPSS, Amos, Stata, and Origin on Bilibili and found that more than 80% of college students learn TCSRS on Bilibili, which shows the characteristic of “fragmentation”. Then, by constructing a model of college students’ OFLE, we clarified the determinants of college students’ OFLE and its mechanisms. Based on the hypothesis testing path of the model, LM, SE, and FTU have a significant positive influence on college students’ OFLE. However, KF does not have a substantial effect on it. Therefore, this section focuses on providing sustainable solutions for the enhancement of college students’ OFLE by proposing resource allocation strategies for online learning platforms, methods to guide college students’ fragmented learning, and countermeasures to changing fragmented learning behaviors so as to better ensure a sustainable and higher-quality online fragmented learning process.

7.1. Build Quality Fragmented Learning Resources and Improve Online Learning Mechanisms

Online courses provide students with a more convenient and flexible way to master the knowledge they need. For some students, online learning is the only way for them to obtain a degree [52]. However, some students need help in judging whether the fragmented learning resources on online learning platforms such as Bilibili are high-quality resources. Therefore, to build high-quality fragmented learning resources, online platforms need to strengthen the access management of teaching video bloggers, improve the quality audit standards of online teaching courses, and adopt college students’ suggestions for courses. At the same time, we need to constantly improve the online fragmented learning mechanisms for college students by adding navigation, mind maps, and hyperlinks to the designs of platforms to make the classification of teaching resources clearer and enhance students’ learning efficiency [14].

7.2. Build a Complete Knowledge System and Improve the Ability of Resource Selection

Research shows that college students generally present the characteristics of fragmented learning time. On the one hand, online teaching lecturers should introduce the knowledge system and context in course videos and pay attention to the coherence of the course chapters and the comprehensibility of the course content to help college students form a complete theoretical framework and build a personalized learning system. On the other hand, due to the limited time for college students to learn in a fragmented way, college students should learn to filter out the resource information that is helpful to their own needs in Baidu and other complicated internet resources and find teaching videos with a high amount of views, strong content consistency, the capability to suit their own needs, and the capability to help them solve practical problems, as this would gradually cultivate college students’ ability to turn fragmented learning into systematic learning and form a personalized knowledge system via “fixed deposit by installments” [16].

7.3. Increase Confidence in Online Learning and Boost LM and SE

LM and SE can significantly and positively affect college students’ OFLE. Learners are the main actors in fragmented learning activities, and LM is essential to ensuring their OFLE. On the one hand, stimulating college students’ LM should help to clarify their learning interests and goals. In online fragmented learning, college students should have a strong sense of self-awareness, find out their interests, formulate reasonable learning goals and plans, improve their self-control ability, and avoid goal deviation [51]. On the other hand, enhancing college students’ SE should help to improve their sense of achievement. College students should adopt a “learning while testing” approach during the learning process. After learning specific knowledge, college students should practice by themselves to test their knowledge mastery. Teachers should provide emotional support by supporting or praising students, alleviating college students’ anxieties about online fragmented learning, helping them to establish an awareness of sustainable online learning, and enhancing their learning confidence [53].

7.4. Strengthen the Concept of Time and Enhance the Ability to FTU

Online fragmented learning has more flexibility in terms of the times and places in which it can take place, providing students with convenient teaching conditions [54]. Using fragmented time consistently for online learning and achieving a better learning effect is something that college students should consider. First, college students should master the ability to manage their fragmented time. For example, if the learning resources regarding the whole course are organized and allocated reasonably, students could make a detailed timetable and schedule tasks for themselves, prioritizing all tasks to ensure they are completed at specific times instead of carrying out unregulated personal study [55]. Secondly, college students should learn to make reasonable use of their fragmented time for learning. For example, they could try to learn when riding the bus or subway and during recess time in between classes [56], as this would encourage learning TCSRS on the online platform anytime, helping to form the habit of using their fragmentation time to develop sustainable online fragmentation learning and maximizing their online fragmented time to improve their learning and achieve their goals.