The Relational Online Learner: Self-Directed Learning, University Belonging, and Student Engagement in Online Higher Education from a Self-Determination Theory Perspective

Abstract

We examined the relationships between self-directed learning skills, university belonging, and student engagement in open online and distance learning environments within the framework of Self-Determination Theory. The research was conducted using a correlational survey model. Data were collected from 696 university students enrolled at a large-scale open, online and distance learning institution in Türkiye. We used three instruments measuring self-directed learning skills, university belonging, and student engagement. The data were analyzed using partial least squares structural equation modeling. The findings indicate that university belonging significantly and positively predicts all dimensions of student engagement. While self-control skills, learning skills, and sustaining the desire to learn significantly explain university belonging, metacognitive awareness and ability to identify sources were found to have no significant effect. Furthermore, it was found that university belonging acts as a significant mechanism in the relationships between self-control skills, learning skills, and sustaining the desire to learn, and student engagement. In this open and distance learning context, student engagement was not explained by individual learning skills alone. University belonging may therefore be one mechanism that partly accounts for the association between self-directed learning and academic participation. The study contributes by testing self-directed learning at the level of its dimensions, by identifying university belonging as a mechanism that partly accounts for its association with engagement, and by situating this model in a large-scale open and distance learning setting.

1. Introduction

Open, online and distance learning (O-ODL) accounts for a growing share of higher education enrolment, and its expansion rests on two specific affordances. The first is flexibility of time and place: students are able to fit their studies around paid work, caregiving and other obligations rather than around a fixed timetable, and they do so without traveling to a campus. The second is access to degree-level study for people whom campus-based provision serves poorly, including adults returning to education after an interruption, those in full-time employment, and students living far from a university or unable to attend regularly because of health or family circumstances (Bozkurt, 2025; Martin & Bolliger, 2018).

These arrangements also move responsibility onto the learner. A physical campus supplies structure that is easy to overlook: timetabled sessions, informal contact with peers and teaching staff, and the visible presence of an academic community. Where that structure is absent, students must initiate, pace and monitor their own study (Broadbent & Poon, 2015). Sustaining participation under these conditions remains a difficulty for institutions and learners alike (Kahu, 2013; Redmond et al., 2018).

Student engagement is a multidimensional construct covering behavioral involvement (effort, participation and persistence), cognitive involvement (depth of processing of course content) and emotional involvement (interest and felt connection to the academic setting) (Kahu & Nelson, 2018; Kuh, 2009). Engagement is harder to sustain online, where interaction is limited and social connectedness is low (Dixson, 2015; Yavuz et al., 2025). The consequences are practical: students who disengage submit fewer assessments, contact instructors and peers less often, are more likely to withdraw from individual courses, and complete their programs at lower rates (Hollister et al., 2022; Pelikan et al., 2021). What sustains engagement in these settings therefore matters for institutional practice as much as for theory.

One learner characteristic repeatedly linked to engagement is self-directed learning (SDL). SDL describes how learners take charge of their own education: diagnosing what they need to learn, setting goals, locating and selecting resources, choosing strategies, and judging the results. It overlaps with self-regulated learning (SRL) without being the same construct. SRL refers to the regulation of cognition, motivation, behavior and metacognitive monitoring within a particular task or study episode, whereas SDL concerns ownership and direction of the learning enterprise as a whole, including decisions taken before and after any single task (Knowles, 1975; Loyens et al., 2008; Zimmerman, 2002). The two terms are often used interchangeably in the online learning literature, which obscures this difference in scope. The focal construct in the present study is SDL; SRL is cited only where the relevant empirical evidence is drawn from that tradition (Edisherashvili et al., 2022; Faza & Lestari, 2025; Luo & Zhou, 2024; Yu, 2023).

SDL is measured as a set of dimensions rather than a single capacity, and there is little reason to assume that every dimension behaves the same way with respect to social outcomes. Some dimensions describe how a learner engages with the learning environment over time: regulating effort and attention, applying study strategies, and sustaining the desire to learn. Others describe competencies that are largely internal or instrumental, such as monitoring one’s own thinking or locating suitable materials. The distinction becomes consequential once the outcome of interest is relational rather than task-based because a skill that helps a student finish a task need not bring that student into contact with the institution.

University belonging (UB) is a student’s sense of being an accepted and valued member of the academic community (Strayhorn, 2018), and it is associated with academic adjustment, motivation and engagement (Allen et al., 2018; Pedler et al., 2022). Belonging warrants particular attention in O-ODL because the conditions that ordinarily produce it are weakened. Incidental contact is rare, communication is largely asynchronous, and membership of the institution is mediated almost entirely through a learning platform. Students in these settings report isolation and distance from the institution (Thomas et al., 2014), which suggests that belonging cannot be treated as an automatic by-product of enrolment.

Self-determination theory (SDT) offers a framework for relating these constructs, provided its terms are mapped onto them explicitly. Within SDT, basic psychological need theory holds that motivation and engagement depend on the satisfaction of needs for autonomy, competence, and relatedness (Vansteenkiste et al., 2020; Ryan & Deci, 2020, 2023). In the model tested here, SDL corresponds to autonomy and competence: learners who set their own direction and manage their study accumulate evidence of their own capability. University belonging corresponds to relatedness. Student engagement is the outcome. Stated this way, the open question is not whether the three needs matter, which is well established, but whether their satisfaction is independent or sequential in a setting where relatedness is structurally constrained.

Research to date supports the separate links but leaves three gaps. First, studies typically treat SDL as a composite score, so it is not known whether its dimensions relate differently to relational outcomes such as belonging. Second, positive associations among SDL, belonging and engagement have been reported (Dai et al., 2022; Yi et al., 2024), but the possibility that belonging accounts for part of the SDL–engagement association has received little direct examination (Chiu, 2022; Yang et al., 2025). Third, reviews of engagement in online learning identify learner-level characteristics such as self-regulated learning strategies, academic self-efficacy and perceived social support as influential without establishing how they operate together (Hu & Xiao, 2025; Ouyang, 2025).

We therefore examined the relationships between SDL, university belonging and student engagement among students enrolled at a large open education institution, using partial least squares structural equation modeling. The study makes three contributions. Theoretically, we test SDL at the level of its dimensions rather than as a composite, which allows for the possibility that dimensions tied to sustained interaction with the learning environment relate to belonging while more solitary cognitive competencies do not. In terms of mechanism, we examine whether university belonging accounts for part of the association between SDL and student engagement rather than treating the two as directly linked. Contextually, we situate the model in an open and distance setting where campus contact is minimal, which makes institutional belonging a matter of deliberate design rather than a by-product of attendance.

2. Literature and Hypotheses

2.1. Self-Directed Learning and University Belonging

According to the SDT perspective, SDL is an important indicator of the fulfillment of students’ needs for autonomy and competence (Ryan & Deci, 2020). It is noted that students who can effectively manage their learning processes are able to develop a stronger psychological bond with the institution and exhibit higher levels of academic adjustment and institutional integration (Kahu & Nelson, 2018; Thomas et al., 2014). Whether every dimension of self-directed learning contributes equally to this integration is less clear. Dimensions involving sustained engagement with the learning environment differ from those that are largely internal, such as monitoring one’s own thinking or locating materials, and the latter need not bring a student into contact with the institution (Broadbent & Poon, 2015; Pelikan et al., 2021). Conversely, students with insufficient SDL skills may experience academic disengagement in online learning environments (Broadbent & Poon, 2015). In this regard, it is believed that SDL skills can positively influence students’ sense of belonging to the university in O-ODL environments (Strayhorn, 2018).

Self-control skills concern regulating effort and attention and holding to study commitments in the absence of an externally imposed timetable. In O-ODL, where pacing is left largely to the learner, students with stronger self-control maintain more regular contact with the platform and with course requirements. Under SDT, repeated successful management of academic demands supplies evidence of competence, and competence satisfaction is theorized to support fuller need satisfaction in the setting where it is exercised (Ryan & Deci, 2020). Broadbent and Poon (2015) report that regulatory strategies predict persistence and academic adjustment in online higher education.

Learning skills concern selecting and adapting methods that suit the learner and the task. Students who can adjust how they study encounter fewer breakdowns in comprehension and are more likely to experience institutional provision as usable rather than obstructive. Kahu and Nelson (2018) position academic self-efficacy as one of the psychosocial constructs through which institutional conditions translate into students’ relationship with their university, which suggests that the capacity to study effectively should relate to how students situate themselves within it.

Metacognitive awareness skills concern monitoring one’s own comprehension and adjusting strategy accordingly. This dimension was expected to relate to belonging through academic adaptation: learners who track their understanding accurately adjust to course demands more successfully, and An et al. (2024) report positive associations between metacognition and learning engagement. The expectation was nonetheless weaker a priori than for the behavioral dimensions because monitoring one’s own thinking is an internal activity that need not bring the learner into contact with the institution or its members.

The ability to identify sources concerns judging the credibility of information and checking it against other sources. It was expected to relate to belonging on the grounds that students who engage critically with course material interact more extensively with what the institution provides. As with metacognitive awareness, this expectation was tentative: evaluating sources is an instrumental activity typically carried out alone, and it is not in itself a form of social or institutional contact.

Sustaining the desire to learn concerns maintaining motivation across a course of study rather than only at its outset. Students who sustain motivation continue to participate after setbacks instead of withdrawing, and continued participation prolongs exposure to peers, teaching staff and institutional routines. SDT treats sustained autonomous motivation as closely bound to relatedness rather than independent of it (Ryan & Deci, 2020; Vansteenkiste et al., 2020), and Yang et al. (2025) report that motivational processes and social belonging operate together in self-directed e-learning.

Accordingly, the following hypotheses have been formulated:

H1a. 

Self-control skills significantly predict UB.

H1b. 

Learning skills significantly predict UB.

H1c. 

Metacognitive awareness skills significantly predict UB.

H1d. 

The ability to identify sources significantly predicts UB.

H1e. 

Sustaining the desire to learn significantly predicts UB.

2.2. UB and Student Engagement

In the context of SDT, belonging refers to the fulfillment of the need for relatedness and directly supports motivation and engagement in learning. It has been noted that students with a strong sense of belonging demonstrate more positive outcomes in terms of academic engagement and learning persistence (Pedler et al., 2022; Wilson et al., 2015). In O-ODL environments, however, the development of a sense of belonging is hindered by the limited nature of face-to-face interaction, and this situation can negatively affect students’ psychological attachment to the institution.

As defined in the Introduction, student engagement comprises behavioral, cognitive and emotional involvement in learning (Kahu & Nelson, 2018), a definition that is used throughout this present study. In higher education, this construct is associated with active participation in classroom activities, cognitive engagement with course content, and sustained academic effort outside of class (Kuh, 2009). Accordingly, students with a strong sense of belonging are expected to demonstrate higher levels of behavioral and cognitive engagement in both face-to-face and distance learning environments.

The four engagement dimensions differ in how socially exposed they are, and belonging was expected to relate to each for partly different reasons. Quiet behavior in the online course covers attending and following instruction. Students who feel accepted are less occupied with the sense of being an outsider and are therefore more available to attend to teaching (Strayhorn, 2018).

Verbal behavior in the online course is the most socially exposed dimension, since contributing to discussion means being seen and evaluated by others. Speaking up presupposes an expectation of a non-hostile reception, which is close to the definition of belonging itself; belonging was therefore expected to relate most directly to this dimension.

Reflecting on online course content involves connecting course material to one’s own life and future. Doing so requires treating the program as personally relevant, and Pedler et al. (2022) link belonging to precisely this kind of investment in study.

Behavior outside the online course covers effort that is not directly required or observed. Because such effort is discretionary, it was expected to depend on the student’s attachment to the institution rather than on compliance alone (Wilson et al., 2015).

Accordingly, the following hypotheses have been formulated:

H2a. 

UB significantly predicts quiet behavior in the online course.

H2b. 

UB significantly predicts verbal behavior in the online course.

H2c. 

UB significantly predicts reflecting on online course content.

H2d. 

UB significantly predicts behavior outside the online course.

2.3. The Mediating Role of University Belonging

Self-directed learning skills have been linked to student engagement; the present model examines whether that link operates through students’ psychological commitment to the university (Kahu & Nelson, 2018). It is noted that students who can effectively manage their learning processes are able to develop a stronger sense of belonging, and that this sense of belonging supports active participation in learning processes (Allen et al., 2018; Pedler et al., 2022; Thomas et al., 2014).

From an SDT perspective, SDL satisfies the needs for autonomy and competence, while UB represents the fulfillment of the need for relatedness (Ryan & Deci, 2020). Within this framework, UB may serve as a psychological mediating mechanism between self-directed learning and student engagement. Because the model specifies self-directed learning and student engagement at the level of their dimensions rather than as composite constructs, the mediating hypotheses are stated dimension by dimension.

Accordingly, the following hypotheses have been formulated:

H3a. 

UB mediates the relationships between self-control skills and the four student engagement dimensions.

H3b. 

UB mediates the relationships between learning skills and the four student engagement dimensions.

H3c. 

UB mediates the relationships between metacognitive awareness skills and the four student engagement dimensions.

H3d. 

UB mediates the relationships between the ability to identify sources and the four student engagement dimensions.

H3e. 

UB mediates the relationships between sustaining the desire to learn and the four student engagement dimensions.

The proposed research model was developed within the framework of SDT to examine the relationships among SDL, UB, and student engagement in O-ODL environments. The hypothesized research model is presented in Figure 1.

As illustrated in Figure 1, the proposed model assumes that SDL dimensions positively contribute to students’ sense of UB, which in turn is associated with the four measured dimensions of student engagement. The dashed path denotes the indirect relationships specified in H3a–H3e, drawn between the two groupings for legibility.

3. Materials and Methods

3.1. Research Model

The purpose of this study is to examine the relationships among self-directed learning, university belonging, and student engagement among students enrolled in O-ODL environments; furthermore, it aims to determine the mediating role of UB in the relationship between SDL and student engagement. Based on this purpose, a structural equation model was specified to estimate the relationships among the five self-directed learning dimensions, university belonging, and the four student engagement dimensions. Structural equation modeling is a statistical method used to test complex theoretical models (Byrne, 2013). Specifically, Partial Least Squares Structural Equation Modeling (PLS-SEM) was selected over Covariance-Based SEM (CB-SEM) due to several distinct theoretical and methodological requirements of the current research design (Hair et al., 2017b). First, while CB-SEM is primarily optimized for strictly confirmatory, theory-testing frameworks aiming to minimize differences between observed and estimated covariance matrices, PLS-SEM operates within a causal-predictive paradigm oriented toward the explained variance (R²) of the endogenous constructs and establishing out-of-sample predictive relevance (Q²) via PLSpredict. Second, the model specified in Figure 1 has a high degree of structural complexity, simultaneously evaluating five distinct dimensions of self-directed learning as exogenous predictors, a unidimensional mediating mechanism (university belonging), and four distinct dimensions of student engagement as separate endogenous outcomes. A model of this size, combining several exogenous constructs with a mediator and multiple endogenous outcomes, is estimated more readily within the variance-based framework of PLS-SEM. Both PLS-SEM and CB-SEM are established approaches to structural model estimation, and this choice reflects the emphasis on explained variance and prediction-oriented assessment rather than a judgment that either method is superior. The study is primarily explanatory, testing theoretically specified relationships, with a secondary predictive component assessed through PLSpredict.

3.2. Data Collection Tools

Data were collected using three measurement instruments, each comprising one or more subscales. First, the “Scale of SDL Skills in Distance Education,” developed by Yılmazer and Kartal (2023), was used to assess students’ SDL. The scale, designed as a 5-point Likert-type scale, comprises 20 items across five subscales: self-control skills, learning skills, metacognitive awareness skills, the ability to identify sources, and sustaining the desire to learn. The instrument was developed with secondary school students learning at a distance. Yılmazer and Kartal (2023) reported Cronbach’s alpha values of 0.78 for self-control skills, 0.77 for learning skills, 0.67 for metacognitive awareness skills, 0.65 for the ability to identify sources, 0.62 for sustaining the desire to learn, and 0.87 for the instrument as a whole. An example item from the learning skills subscale is “I can identify the learning method that is easiest for me during distance education.” All items appear in Appendix A.

In this study, the 5-point Likert-type “Short College Belonging Scale” was used to determine students’ levels of UB and to identify the relationship between other scales. The scale was developed by Li (2023) and adapted into Turkish by Ekşi and Ay (2024). The instrument is unidimensional and comprises four items, of which the negatively worded items were reverse-scored before analysis. Ekşi and Ay (2024) reported an overall Cronbach’s alpha of 0.85. An example item is “I feel that I belong at my university.” All items appear in Appendix A.

Finally, we used an instrument developed by Mazer (2012) and adapted into Turkish by Yıldırım et al. (2018) to determine students’ levels of participation in online courses. The scale comprises 13 items across four dimensions: quiet behavior in the online course, verbal behavior in the online course, reflecting on online course content, and behavior outside the online course. Yıldırım et al. (2018) reported Cronbach’s alpha values of 0.81 for quiet behavior in the online course, 0.91 for verbal behavior in the online course, 0.70 for reflecting on online course content and 0.77 for behavior outside the online course. An example item from the verbal behavior subscale is “I contribute to class discussions by expressing my ideas.” All items appear in Appendix A.

3.3. Sampling Procedure and Data Collection Process

This study was conducted online with students of the Faculty of Open Education at Anadolu University, a large-scale institution with approximately one million students enrolled each semester (Bozkurt, 2025). The data collection process was carried out through the Anadolu University e-Campus learning management system in accordance with the principles of confidentiality and voluntariness. The study was approved by the Scientific Research and Publication Ethics Board for Social and Human Sciences of Anadolu University (application no. 161840, 5 June 2026).

The survey consisted of two sections: demographic questions and the items of the three measurement instruments. The demographic questions included gender, age, education level, grade level, and frequency of e-campus use. The second section consisted of 38 questions: the 37 instrument items and one attention-check item used to identify careless responding. The attention-check item was placed among the scale items and read “If you are reading this item, please select ‘undecided’.” A total of 743 students submitted a complete response. Forty-seven respondents did not select the instructed option and were treated as careless responders, leaving 696 cases for analysis. For sample selection, the convenience sampling method (Golzar et al., 2022), which relies on easily accessible and voluntary participants, was used. Demographic information for the students included in the sample is provided in

Table 1. Demographic information on participants (N = 696).

Variable	f	%
Gender
Female	469	67.4%
Male	227	32.6%
Age Group
18–20	55	7.9%
21–25	102	14.7%
26–30	88	12.6%
31–35	58	8.3%
36–40	70	10.1%
41–45	85	12.2%
46–50	87	12.5%
51–55	57	8.2%
56–60	37	5.3%
61 and above	57	8.2%
Education Level
Associate Degree	300	43.1%
Bachelor’s Degree	396	56.9%
Grade Level
First year	289	41.5%
Second year	211	30.3%
Third year	81	11.6%
Fourth year	115	16.5%
Frequency of E-Campus Use
Never (1)	20	2.9%
Rarely (2)	64	9.2%
Sometimes (3)	146	21.0%
Often (4)	148	21.3%
Always (5)	318	45.7%
Total	696	100%

Note. f = frequency.

.

3.4. Validity and Reliability Analyses of Scales

Before testing the research model, construct validity and reliability were assessed, covering internal consistency, convergent validity, and discriminant validity. Internal consistency was assessed using Cronbach’s alpha and composite reliability (CR), and convergent validity was evaluated using factor loadings and average variance extracted (AVE). Thresholds were set at factor loadings ≥ 0.70, Cronbach’s alpha and CR ≥ 0.70, and AVE ≥ 0.50 (Fornell & Larcker, 1981; Hair et al., 2017a). Results are shown in

Table 2. Reliability and convergent validity.

Scale	Dimensions	Items	Factor Loading	Cronbach’s Alpha	Reliability Coefficient (ρA)	Composite Reliability (ρC)	AVE
Self-Directed Learning Skills Scale	Self-control skills	SCS1	0.82	0.75	0.77	0.84	0.58
		SCS2	0.75
		SCS4	0.69
		SCS5	0.76
	Learning skills	LS1	0.79	0.78	0.80	0.85	0.54
		LS2	0.81
		LS3	0.72
		LS4	0.76
		LS5	0.55
	Metacognitive awareness skills	MAS1	0.67	0.76	0.79	0.85	0.58
		MAS2	0.81
		MAS3	0.83
		MAS4	0.73
	Ability to identify sources	AIS1	0.74	0.70	0.84	0.82	0.60
		AIS2	0.87
		AIS3	0.70
	Sustaining the Desire to Learn	SDTL1	0.80	0.70	0.71	0.83	0.62
		SDTL2	0.77
		SDTL3	0.80
Student Engagement Scale	Quiet Behavior in the Online Course	QBC1	0.91	0.84	0.85	0.90	0.76
		QBC2	0.91
		QBC3	0.78
	Verbal Behavior in the Online Course	VBC1	0.80	0.80	0.80	0.88	0.71
		VBC2	0.90
		VBC3	0.82
	Reflecting on Online Course Content	RCC1	0.72	0.72	0.73	0.84	0.65
		RCC2	0.86
		RCC3	0.82
	Behavior Outside the Online Course	BOC1	0.89	0.67	0.80	0.81	0.59
		BOC2	0.72
		BOC4	0.67
University Belonging Scale	University Belonging	UB1	0.84	0.76	0.79	0.85	0.59
		UB2	0.83
		UB3	0.61
		UB4	0.77

.

An examination of Table 2 reveals that all scales used are psychometrically adequate and reliable. Factor loadings range from 0.55 to 0.91, with the vast majority exceeding 0.70. This indicates that the items generally represent their respective dimensions strongly. Cronbach’s alpha values ranged from 0.67 to 0.84 and composite reliability values from 0.71 to 0.90. All constructs met the 0.70 criterion except Behavior Outside the Online Course, whose alpha (0.67) fell marginally below it. Hair et al. (2011) and Hair et al. (2017a) recommend that items with outer loadings below 0.40 be removed, whereas items loading between 0.40 and 0.70 should be removed only where removal raises composite reliability or AVE above the recommended thresholds of 0.70 and 0.50, respectively. Upon examination of the scale, it was determined that the factor loading of item SCS3 was 0.12, which fell below the recommended threshold of 0.40. Therefore, this item was removed from the model. AVE and CR values were examined for the remaining items. Following the removal of this item, all other indicators were retained. Five load below the preferred 0.70 threshold: SCS4 (λ = 0.69), LS5 (λ = 0.55), MAS1 (λ = 0.67), BOC4 (λ = 0.67) and UB3 (λ = 0.61). These were deliberately preserved in the model based on the robust psychometric guidelines outlined by Hair et al. (2017a). According to these established criteria, indicators with factor loadings between 0.40 and 0.70 should only be eliminated if their removal leads to a substantial increase in Composite Reliability (CR) or Average Variance Extracted (AVE) above the recommended thresholds. In this model, the latent constructs for learning skills and university belonging both exhibited strong structural health, with CR values (0.85 for both constructs) and AVE values (0.54 for learning skills; 0.59 for UB) safely exceeding the required thresholds of 0.70 and 0.50, respectively. Consequently, retaining these items is methodologically justified as it preserves the theoretical content validity and domain coverage of the scales without compromising the overall convergent validity or statistical reliability of the measurement model. Convergent validity was supported, as every construct satisfied both the AVE ≥ 0.50 and the CR ≥ 0.70 criteria (Fornell & Larcker, 1981). It is noted that when the conditions AVE ≥ 0.50 and CR ≥ 0.70 are met, such items can be retained in the scale without compromising construct validity (Hair et al., 2017a). Similarly, Fornell and Larcker (1981) emphasize that convergent validity should be assessed not only based on individual factor loadings but also by considering the structure’s overall ability to explain variance. Thus, the scale’s content integrity was preserved, and its construct validity was statistically supported. The retention and removal decisions are summarized in

Table 3. Item retention and removal decisions.

Construct	Item	Loading	Decision	Rationale
Self-control skills	SCS3	0.12	Removed	Below the 0.40 criterion
Self-control skills	SCS4	0.69	Retained	CR = 0.84, AVE = 0.58 already above thresholds
Learning skills	LS5	0.55	Retained	CR = 0.85, AVE = 0.54; preserves content coverage
Metacognitive awareness skills	MAS1	0.67	Retained	CR = 0.85, AVE = 0.58 above thresholds
Behavior outside the online course	BOC3	0.68	Removed	High cross-loadings on other constructs
Behavior outside the online course	BOC4	0.67	Retained	CR = 0.81, AVE = 0.59 above thresholds
University belonging	UB3	0.61	Retained	CR = 0.85, AVE = 0.59 above thresholds

Note. All items not listed loaded at or above 0.70 and were retained. Removal decisions were made on statistical grounds and then checked against content coverage; no item was removed solely to raise CR or AVE.

.

The analysis revealed that Cronbach’s alpha values ranged from 0.67 to 0.84, ρA values ranged from 0.71 to 0.85, and ρC values ranged from 0.81 to 0.90. With the exception of Cronbach’s alpha for behavior outside the online course (0.67), these coefficients exceeded the recommended threshold of 0.70. Because composite reliability is generally preferred to Cronbach’s alpha in PLS-SEM, and because ρA (0.80), ρC (0.81), and AVE (0.59) for this construct all exceeded their respective thresholds, this construct was retained (Hair et al., 2017a). Additionally, AVE values ranged from 0.54 to 0.76. An AVE value above 0.50 indicates that convergent validity is established and that the items collectively represent a common structure (Fornell & Larcker, 1981). Therefore, it can be concluded that the structures in the model provide reliable and valid measurements.

3.5. Discriminant Validity

To strengthen the discriminant validity of the measurement model, the cross-loadings of the items were first examined. Following established guidance (Hair et al., 2017a; Henseler et al., 2015), the loading of an item on its target factor is expected to be significantly higher than its loading on other factors. Otherwise, the item measures multiple constructs simultaneously and creates conceptual ambiguity (Hair et al., 2017a). Henseler et al. (2015) note that high cross-loadings weaken discriminant validity and prevent constructs from being sufficiently distinguished from one another. For this reason, item BOC3, which exhibited high loadings on other factors as well, was removed from the study to enhance the theoretical clarity and psychometric quality of the measurement model. This process ensured that the constructs were distinguished from one another more clearly and strengthened the validity of the measurement model. Similarly, Fornell and Larcker (1981) emphasize that, to ensure discriminant validity, each item must represent only its own structure. Consequently, the removal of items showing cross-loadings was implemented as a methodological requirement recommended in the literature on scale development and structural equation modeling.

The Fornell–Larcker criterion was used to assess whether the constructs were distinguishable from one another and thus to test their discriminant validity. In this approach, the square root of the AVE for each construct is compared with the correlations between that construct and the other constructs. According to the Fornell–Larcker criterion, a construct’s discriminant validity is considered established if the square root of its AVE is higher than the correlations of that construct with all other constructs in the model (Fornell & Larcker, 1981). The results of the analysis are presented in

Table 4. Discriminant validity results (Fornell–Larcker criterion).

	AIS	BOC	UB	LS	MAS	QBC	RCC	SCS	SDTL	VBC
AIS	0.77
BOC	0.44	0.77
UB	0.26	0.31	0.77
LS	0.34	0.45	0.53	0.73
MAS	0.42	0.55	0.42	0.66	0.76
QBC	0.30	0.42	0.35	0.33	0.43	0.87
RCC	0.39	0.51	0.35	0.48	0.42	0.30	0.80
SCS	0.47	0.58	0.54	0.71	0.70	0.41	0.48	0.76
SDTL	0.34	0.45	0.50	0.64	0.56	0.40	0.32	0.65	0.79
VBC	0.23	0.30	0.29	0.20	0.27	0.53	0.23	0.32	0.37	0.84

.

Upon examining the discriminant validity results based on the Fornell–Larcker criterion presented in Table 4, it is observed that the square root values of AVE along the diagonal are higher than the correlation coefficients between the respective constructs and other constructs. Accordingly, the AVE square root values for the ability to identify sources (0.77), behavior outside the online course (0.77), UB (0.77), learning skills (0.73), metacognitive awareness skills (0.76), quiet behavior in the online course (0.87), reflecting on online course content (0.80), self-control skills (0.76), sustaining the desire to learn (0.79), and verbal behavior in the online course (0.84) are all greater than the correlation values in the corresponding rows and columns. Although the highest correlations were observed between learning skills and self-control skills (0.71), between metacognitive awareness skills and self-control skills (0.70), and between learning skills and metacognitive awareness skills (0.66), these values remain below the AVE square root values of the respective constructs. The margin is narrowest for learning skills, whose AVE square root (0.73) exceeds its correlation with self-control skills (0.71) by a small amount. This finding indicates that the constructs included in the model are sufficiently distinguished from one another. Overall, it can be stated that the measurement model meets the discriminant validity criterion according to the Fornell–Larcker test and is therefore acceptable in this regard (Fornell & Larcker, 1981).

4. Results

This section presents findings regarding the relationships among self-directed learning, UB, and student engagement among students enrolled in O-ODL environments. Partial Least Squares Structural Equation Modeling (PLS-SEM) was used to assess the relationships among the research variables. PLS-SEM was chosen for the reasons set out in Section 3.1: the emphasis on explained variance, the number of endogenous constructs in the model, and the secondary predictive assessment carried out through PLSpredict. The model developed for this study is presented in Figure 2.

The PLS algorithm was executed for structural evaluation, and path coefficients, coefficients of determination (R²), and effect sizes (f²) were calculated. The predictive relevance of the model was assessed via PLSpredict, from which the Q² values were obtained. The significance of the path coefficients was evaluated through the bootstrapping method with 5000 subsamples. The VIF, R², f², and Q² values—used to assess the structural adequacy of the model and the validity of the relationships between variables—are presented in

Table 5. Coefficients of the structural model shown in Figure 2.

Predictor	Outcome	VIF	f2	R2	Q2 Predict
SCS	UB	2.89	0.04	0.35	0.34
LS		2.48	0.03
MAS		2.27	0.00
AIS		1.31	0.00
SDTL		1.95	0.03
UB	QBC	1.00	0.14	0.12	0.13
UB	VBC	1.00	0.09	0.08	0.08
UB	RCC	1.00	0.14	0.13	0.16
UB	BOC	1.00	0.11	0.10	0.17

.

An examination of Table 5 reveals that the VIF values for the variables are below the threshold of 5, indicating that problematic collinearity was not detected (Hair et al., 2017a).

For the effect size coefficient f², values above 0.02 are considered small, values of 0.15 and above are considered moderate, and values of 0.35 and above are considered large (Cohen, 1988). Based on the results in Table 5, it was observed that among the variables predicting the UB variable, self-control skills (f² = 0.04), learning skills (f² = 0.03), and sustaining the desire to learn (f² = 0.03) have a low level of effect. In contrast, it was observed that the effect sizes of the metacognitive awareness skills (f² = 0.00) and the ability to identify sources (f² = 0.00) variables on UB are quite limited. When evaluated in terms of student engagement’s sub-dimensions, it can be stated that there is a low-level effect on the quiet behavior in the online course (f² = 0.14), reflecting on online course content (f² = 0.14), behavior outside the online course (f² = 0.11), and verbal behavior in the online course (f² = 0.09) variables. Overall, the f² values indicate that the effects in the model are mostly at a low level.

An examination of the R² values obtained from the model reveals that the explained variance for the UB variable is 0.35. This result indicates that the self-control skills, learning skills, metacognitive awareness skills, the ability to identify sources, and sustaining the desire to learn together explain 35% of the variance in UB. In terms of student engagement’s sub-dimensions, the explained variance values were calculated as 0.12 for quiet behavior in the online course, 0.08 for verbal behavior in the online course, 0.13 for reflecting on online course content, and 0.10 for behavior outside the online course. These findings indicate that the model has moderate explanatory power for the UB variable and more limited explanatory power for the sub-dimensions. Furthermore, the fact that all VIF values fall within acceptable limits indicates that there is no multicollinearity issue in the model.

The fact that the Q² predicted values are greater than zero indicates that the research model possesses predictive power (Hair et al., 2017a). Upon examining Table 5, it can be seen that the Q² predict value for UB is 0.34 and that the model offers positive predictive power for the UB variable. In terms of sub-dimensions, the Q² predict values were calculated as 0.13 for quiet behavior in the online course, 0.08 for verbal behavior in the online course, 0.16 for reflecting on online course content, and 0.17 for behavior outside the online course. The fact that all these values are greater than zero indicates that the model possesses predictive validity for the relevant variables. Overall, the research model demonstrates a more pronounced predictive power for the UB variable and a lower but positive predictive power for the sub-dimensions. The results regarding direct effects in the research model are presented in

Table 6. Path coefficients and significance levels for hypotheses regarding direct effects.

Hypothesis	Path	β	SD	t	p	Result
H1a	SCS -> UB	0.27	0.06	4.66	<0.001	Accepted
H1b	LS -> UB	0.23	0.06	3.84	<0.001	Accepted
H1c	MAS -> UB	0.03	0.06	0.55	0.58	Rejected
H1d	AIS -> UB	0.00	0.04	0.04	0.97	Rejected
H1e	SDTL -> UB	0.20	0.05	4.12	<0.001	Accepted
H2a	UB -> QBC	0.35	0.04	9.12	<0.001	Accepted
H2b	UB -> VBC	0.29	0.04	7.64	<0.001	Accepted
H2c	UB -> RCC	0.36	0.04	8.94	<0.001	Accepted
H2d	UB -> BOC	0.31	0.04	8.47	<0.001	Accepted

Note. β = standardized path coefficient; t = t-statistic; p < 0.001.

.

Examining Table 6 reveals that the self-control skills variable has a significant effect on UB (β = 0.27; t = 4.66; p ≤ 0.001), the learning skills variable has a significant effect on UB (β = 0.23; t = 3.84; p ≤ 0.001), and sustaining the desire to learn on UB (β = 0.20; t = 4.12; p ≤ 0.001). In contrast, it was found that the metacognitive awareness skills (β = 0.03; t = 0.55; p = 0.58) and the ability to identify sources (β = 0.00; t = 0.04; p = 0.97) variables do not have a statistically significant effect on UB. Additionally, UB was found to have a significant effect on quiet behavior in the online course (β = 0.35; t = 9.12; p ≤ 0.001), verbal behavior in the online course (β = 0.29; t = 7.64; p ≤ 0.001), reflecting on online course content (β = 0.36; t = 8.94; p ≤ 0.001), and behavior outside the online course (β = 0.31; t = 8.47; p ≤ 0.001). Based on these findings, it is observed that all hypotheses except H1c and H1d are statistically supported. Finally, the results regarding indirect and mediating effects in the research model are presented in

Table 7. Path coefficients and significance levels for hypotheses regarding indirect effects.

Hypothesis	Path	β	SD	t	p	Result
H3a	SCS → QBC	0.10	0.02	4.01	<0.001	Accepted
H3a	SCS → VBC	0.08	0.02	3.83	<0.001	Accepted
H3a	SCS → RCC	0.10	0.02	4.10	<0.001	Accepted
H3a	SCS → BOC	0.08	0.02	3.84	<0.001	Accepted
H3b	LS → QBC	0.08	0.02	3.73	<0.001	Accepted
H3b	LS → VBC	0.07	0.02	3.71	<0.001	Accepted
H3b	LS → RCC	0.08	0.02	3.29	<0.001	Accepted
H3b	LS → BOC	0.07	0.02	3.71	<0.001	Accepted
H3c	MAS → QBC	−0.01	0.02	0.55	0.59	Rejected
H3c	MAS → VBC	−0.01	0.02	0.54	0.59	Rejected
H3c	MAS → RCC	−0.01	0.02	0.54	0.59	Rejected
H3c	MAS → BOC	−0.01	0.02	0.54	0.59	Rejected
H3d	AIS → QBC	0.00	0.01	0.04	0.97	Rejected
H3d	AIS → VBC	0.00	0.01	0.04	0.97	Rejected
H3d	AIS → RCC	0.00	0.01	0.04	0.97	Rejected
H3d	AIS → BOC	0.00	0.01	0.04	0.97	Rejected
H3e	SDTL → QBC	0.07	0.02	3.50	<0.001	Accepted
H3e	SDTL → VBC	0.06	0.02	3.28	<0.001	Accepted
H3e	SDTL → RCC	0.07	0.02	3.73	<0.001	Accepted
H3e	SDTL → BOC	0.06	0.02	3.46	<0.001	Accepted

Note. β = standardized path coefficient; t = t-statistic; p < 0.001. Because no direct paths were specified from the self-directed learning dimensions to the student engagement dimensions, the total effect of each dimension on each outcome is equal to the indirect effect reported here.

.

The total indirect effects presented in Table 7 were examined to assess the mediating role of UB. Self-control skills had significant indirect effects through UB on quiet behavior in the online course (β = 0.10, t = 4.01, p < 0.001), verbal behavior in the online course (β = 0.08, t = 3.83, p < 0.001), reflecting on online course content (β = 0.10, t = 4.10, p < 0.001) and behavior outside the online course (β = 0.08, t = 3.84, p < 0.001). Learning skills showed the same pattern across all four outcomes (β = 0.07–0.08, t = 3.29–3.73, p < 0.001), as did sustaining the desire to learn (β = 0.06–0.07, t = 3.28–3.73, p < 0.001). In contrast, neither metacognitive awareness skills (β = −0.01, t = 0.54–0.55, p = 0.59) nor the ability to identify sources (β = 0.00, t = 0.04, p = 0.97) produced statistically significant indirect effects on any outcome. H3a, H3b, and H3e were therefore supported for all four engagement dimensions, whereas H3c and H3d were not supported.

5. Discussion

In this study, the relationships between SDL, UB, and student engagement in O-ODL environments were examined within the framework of SDT. The findings obtained in this section were discussed considering the relevant literature and theoretical framework. The findings of this study indicate that UB is a consistent and significant predictor of all dimensions of student engagement. This result can be explained, particularly within the context of SDT, through the need for relatedness. According to Ryan and Deci (2020), relatedness is one of the fundamental psychological needs that supports an individual’s motivation and behavioral persistence. The development of psychological attachment to the institution among students in O-ODL environments does not limit participation to merely attending online courses or completing assignments. Rather, it supports students’ deeper cognitive and behavioral engagement in the academic process.

This finding is consistent with Strayhorn’s (2018) argument, which treats belonging as a motivational factor shaping students’ responses to academic demands. It also aligns with Kahu and Nelson’s (2018) educational interface model, which locates student engagement at the intersection of institutional structures and individual student characteristics and treats institutional belonging as one of the psychosocial interfaces between the structural conditions of the learning environment and student engagement. Furthermore, these findings generally align with the studies by Chiu (2022), Yi et al. (2024), and Adlington et al. (2024), which demonstrate that UB serves as a central psychological mechanism supporting student participation in online learning environments, as well as with the studies by Sun et al. (2023), which demonstrate that SDL strengthens online participation. However, it partially diverges from the study by An et al. (2024), which indicates that metacognition is positively associated with learning participation, in that metacognitive awareness does not significantly predict belonging in the current study. This divergence may reflect the setting, since the conditions under which belonging forms in online learning differ from those in face-to-face education. Limited physical proximity, a reduction in spontaneous social interactions, and the fact that communication is often asynchronous can make students’ perception of belonging to the institution more critical.

When examining which dimensions of SDL predict UB, a more nuanced picture emerges. The findings indicate that self-control skills, learning skills, and sustaining the desire to learn are positively associated with UB. In contrast, metacognitive awareness and the ability to identify sources were found to have no significant effect on belonging. This situation can be explained through a mechanism-based approach within the framework of SDT (Ryan & Deci, 2020). Self-control skills, learning skills, and sustaining the desire to learn may support students in adapting more effectively to academic processes. These skills may also contribute to reducing feelings of inadequacy in online learning environments. Students who can manage their learning processes, meet course expectations, and sustain their motivation may, over time, develop a stronger perception of their own academic competence. According to Self-Determination Theory (Ryan & Deci, 2020), a sense of competence is not solely related to individual success. It can also support the individual in forming a stronger psychological bond with the academic community they are part of (Thomas et al., 2014). In this context, academic success can be viewed not merely as a cognitive outcome but also as a relational process that supports the student’s emergence as a legitimate and active participant in academic life. This interpretation aligns with Kahu and Nelson’s (2018) approach, which emphasizes that academic self-efficacy supports psychological belonging, and with Broadbent and Poon’s (2015) findings, which demonstrate that self-control skills enhance persistence and academic adjustment in online learning.

It is noteworthy that metacognitive awareness skills and the ability to identify sources do not have a significant impact on a sense of belonging. These skills relate more closely to cognitive and task-management processes such as planning study sessions, monitoring comprehension, and evaluating information sources. While metacognitive awareness focuses on an individual’s ability to monitor their own thinking processes, the items measuring the ability to identify sources concern judging the credibility of information and checking it against other sources (Appendix A). Indeed, course materials in O-ODL systems are deliberately designed around individual learning principles (Yavuz et al., 2020); thus, the skills required to access and manage them remain largely task-centric and solitary. Such skills may support the completion of learning tasks. However, they may not directly ensure that the student feels valued and accepted within the institutional structure. Indeed, Strayhorn (2018) and Pedler et al. (2022) emphasize that a sense of belonging is a socially constructed experience with emotional foundations. Similarly, Pelikan et al. (2021) and Hollister et al. (2022) point out that cognitive and self-regulatory skills in online learning environments may not automatically translate into social integration. Therefore, it can be argued that technical or cognitive competencies alone may not be sufficient to generate psychological belonging. This finding suggests that not all dimensions of SDL have the same level of impact on psychosocial outcomes. This distinction appears theoretically relevant in the O-ODL context because it indicates that the relational dimensions of cognitive self-regulation and institutional identity must be distinguished from one another.

One of the most significant theoretical findings of this study is the mediating role of UB. The findings indicate that the relationships between self-control skills, learning skills, sustaining the desire to learn, and the four dimensions of student engagement are mediated by UB. This result addresses a question that has received limited attention in the O-ODL literature, namely, how self-directed learning relates to student engagement. Significant indirect associations through university belonging were found for self-control skills, learning skills, and sustaining the desire to learn. Students scoring higher on these dimensions reported a stronger sense of belonging, which was in turn associated with higher engagement.

This situation can be explained by the gradual integration of psychological needs within the framework of Self-Determination Theory (Ryan & Deci, 2020). Self-control skills and learning skills can support students’ need for competence. A strengthened sense of competence can increase the student’s sense of belonging to the institution. This process can contribute to the fulfillment of the need for relatedness. Consequently, students may move beyond being merely individuals who fulfill academic tasks. Instead, they can participate in the learning process with a more internalized motivation (Chiu, 2022; Ryan & Deci, 2020). In this regard, UB is better understood as part of the pathway between individual learning capacity and academic participation than as a simple outcome of SDL. This interpretation aligns with the approach in Yang et al. (2025), which frames social belonging processes in online learning environments as a motivational mediating mechanism. Similarly, Yi et al. (2024) note that the relationship between self-directed behaviors and academic outcomes in online learning environments is shaped by students’ perceptions of institutional commitment. The current study makes a significant contribution to this literature. The findings demonstrate that UB is not merely a relational variable in the online participation model.

Although the relationships are statistically significant, the effect sizes are generally found to be low. This is an expected result given the contextual characteristics of the study. Large-scale O-ODL environments have highly heterogeneous student groups. In these environments, students differ in terms of age, professional background, geographic location, and prior educational experiences. The fact that the current sample spans a wide age range and includes students from different grade levels also reflects this diversity. In such environments, it should not be expected that a single psychological or behavioral variable can fully explain student engagement. These effect sizes are small in Cohen’s (1988) terms and are interpreted as such here. University belonging is therefore best read as one of several factors associated with student engagement rather than as a dominant explanatory mechanism. This is because student engagement is not influenced solely by individual learning skills. Many factors, such as technological access, living conditions, work obligations, family responsibilities, and institutional support, can also shape this process (Kahu & Nelson, 2018).

The small effect sizes reported above are consistent with Broadbent and Poon (2015) and Dixson (2015), who noted that participation in online higher education is a context-sensitive construct shaped by multiple factors. Belonging and self-directed learning were nonetheless associated with student participation in this setting, alongside the many other factors that shape it. In this regard, the findings are significant in that they reveal realistic mechanisms operating in large-scale open and distance education contexts, rather than in more controlled and homogeneous samples. Taken together, the findings are consistent with a sequential rather than a parallel reading of basic psychological needs in this setting. Self-directed learning behaviors relate to autonomy and competence, university belonging relates to relatedness, and belonging accounts for part of the association between self-directed learning and student engagement. Because the design is cross-sectional, this ordering is offered as an interpretation to be tested rather than as a demonstrated causal sequence.

6. Conclusions

We examined the relationships between self-directed learning skills, UB, and student engagement in O-ODL environments within the framework of SDT. Overall, the findings indicate that student engagement in this context was not explained by individual SDL capacities alone. The relationship between students’ ability to manage their own learning and their academic engagement becomes clearer when considered alongside their psychological attachment to the institution. The absence of physical presence in O-ODL environments necessitates that a sense of institutional community be consciously constructed rather than arising spontaneously. Models explaining student participation in online education may be theoretically incomplete if they fail to account for the social and relational dimensions of the learner’s experience.

From a theoretical perspective, this study contributes to the existing literature in several ways. First, the findings indicate that the fundamental psychological needs outlined in SDT do not operate independently of one another in online learning environments. In this context, the needs for autonomy, competence, and relatedness can be viewed as sequential and complementary processes. Self-control skills, learning skills, and sustaining the desire to learn can support students’ needs for autonomy and competence. However, meeting these needs may not directly translate into student engagement unless the need for relatedness is supported through institutional belonging.

The second theoretical contribution of this study is the distinction revealed by the non-meaningful pathways regarding metacognitive awareness skills and the ability to identify sources. These dimensions are primarily cognitive and instrumental in nature. They can contribute to managing learning tasks. However, they are not directly related to the social ecology of the learning institution. Therefore, they may not, on their own, generate the sense of acceptance, being valued, and institutional commitment that belonging requires. This distinction suggests that cognitive self-management and motivation-based institutional identification are separable processes. This situation may help explain why some online learners, despite possessing high levels of self-directed learning skills, may eventually drift away from participation over time. These students may possess the cognitive tools necessary for learning. However, they may lack the psychological foundations that make sustained participation meaningful.

Third, this study positions UB not merely as a supporting variable in the context of online learning, but as the mediating variable examined in the relationship between SDL and student engagement. This contribution proposes a more holistic model that does not reduce student engagement to individual skill levels. Taken together, the findings point to both individual learning capacity and institutional psychological ties as relevant to engagement in settings of this kind.

These findings offer important practical implications for institutions that design, manage, and support O-ODL environments. One of the study’s key messages is that while investments in technological infrastructure and content delivery are necessary, they may not be sufficient on their own. To sustain student engagement, there is also a need for deliberate practices that strengthen a sense of belonging among online learners.

For universities operating in large-scale online learning environments, the findings point to the potential value of learning management systems that do more than deliver content. Such systems can be structured to support social interaction among students and to foster a sense of academic community. Peer interaction, online mentoring, collaborative learning activities, and structured opportunities for students to contribute to the academic community are crucial in this regard. Such initiatives can help create the psychological conditions necessary for student engagement.

Instructional designers and academic support units may also consider the motivational and relational dimensions of the learner’s experience alongside the cognitive and procedural ones. Students who appear technically ready to learn may feel invisible or excluded within the institutional structure. Over time, this can increase the risk of disengagement and difficulty in continuing their education. Therefore, for open and distance education institutions serving large and heterogeneous student groups, the fundamental challenge is to design a sense of belonging in a scalable manner. Making the experience of belonging to an academic community accessible to all students, rather than only to those who are technologically confident, is therefore a design consideration for these environments.

This study has certain limitations. These limitations help define the scope of the results’ interpretation and suggest new directions for future research. The study was conducted within the context of a single large-scale O-ODL institution in Türkiye. The findings, therefore, describe students within this Turkish open and distance learning context. They should not be generalized to other open education institutions, to campus-based higher education, or to smaller online programs operating with different technological infrastructures. Convenience sampling was used in the study. This method offers a practical advantage in terms of accessing large online student groups. However, there is a possibility of systematic self-selection bias in the sample. Additionally, the study’s cross-sectional design does not allow for causal inferences. While the structural model presents theoretically meaningful relationships, it cannot explain how self-directed learning skills develop, how a sense of belonging forms over time, or how student engagement changes. Participants also varied widely in age, grade level, and frequency of e-campus use. Subgroup differences were not modeled here because the study was designed to estimate the overall structural relationships rather than to compare groups, and because several subgroups were too small for stable multi-group estimation. Future research should establish measurement invariance and then compare the model across demographic and usage-based groups.

Another limitation is that all variables were measured using self-report scales. This may increase the risk of common-method bias. Procedural remedies were limited to guaranteeing anonymity, emphasizing voluntary participation, and including an attention-check item; no statistical test for common-method variance was performed, so the associations reported here may be inflated by shared method variance and should be read with that in mind. Furthermore, self-report data may not fully reflect actual behavioral patterns in online learning environments. Future research could examine how belonging and engagement develop over the course of an academic term using longitudinal or experiential sampling designs. Furthermore, the proposed model could be tested in different cultural contexts, blended learning environments, and various open education systems.

Future studies may also use experimental or quasi-experimental designs. One feasible option would be to offer a belonging-focused orientation program to one cohort but not another and to measure engagement before and after, which would indicate whether changes in belonging precede changes in engagement. This would allow for stronger causal inferences regarding the relationships between SDL, a sense of belonging, and student engagement. It may also be beneficial to include additional psychological mediating variables in the model, such as academic self-efficacy, emotional well-being, and perceived social support. Furthermore, AI-supported online learning environments present an important area for future research. The increasing integration of adaptive and intelligent technologies into distance education systems necessitates an examination of the effects of these technologies on belonging and participation (Üstün et al., 2026).

In conclusion, the findings suggest that the opportunities offered by O-ODL are not realized through access to technology or individual learning skills alone. In settings of this kind, sustained participation may also depend on the psychological bond students form with their academic community. Students should be viewed not merely as users of an educational platform, but as members of an institution where their participation is recognized and valued. Therefore, creating conditions that strengthen psychological belonging in the design of online learning environments can be considered one of the most important investments distance education institutions can make for students’ long-term academic success and well-being.